Daily Papers

We present a simple approach which can turn a ViT encoder into an efficient video model, which can seamlessly work with both image and video inputs. By sparsely sampling the inputs, the model is able to do training and inference from both inputs. The model is easily scalable and can be adapted to large-scale pre-trained ViTs without requiring full finetuning. The model achieves SOTA results and the code will be open-sourced.

Scale is the primary factor for building a powerful foundation model that could well generalize to a variety of downstream tasks. However, it is still challenging to train video foundation models with billions of parameters. This paper shows that video masked autoencoder (VideoMAE) is a scalable and general self-supervised pre-trainer for building video foundation models. We scale the VideoMAE in both model and data with a core design. Specifically, we present a dual masking strategy for efficient pre-training, with an encoder operating on a subset of video tokens and a decoder processing another subset of video tokens. Although VideoMAE is very efficient due to high masking ratio in encoder, masking decoder can still further reduce the overall computational cost. This enables the efficient pre-training of billion-level models in video. We also use a progressive training paradigm that involves an initial pre-training on a diverse multi-sourced unlabeled dataset, followed by a post-pre-training on a mixed labeled dataset. Finally, we successfully train a video ViT model with a billion parameters, which achieves a new state-of-the-art performance on the datasets of Kinetics (90.0% on K400 and 89.9% on K600) and Something-Something (68.7% on V1 and 77.0% on V2). In addition, we extensively verify the pre-trained video ViT models on a variety of downstream tasks, demonstrating its effectiveness as a general video representation learner. The code and model is available at https://github.com/OpenGVLab/VideoMAEv2.

We present an architecture and a training recipe that adapts pre-trained open-world image models to localization in videos. Understanding the open visual world (without being constrained by fixed label spaces) is crucial for many real-world vision tasks. Contrastive pre-training on large image-text datasets has recently led to significant improvements for image-level tasks. For more structured tasks involving object localization applying pre-trained models is more challenging. This is particularly true for video tasks, where task-specific data is limited. We show successful transfer of open-world models by building on the OWL-ViT open-vocabulary detection model and adapting it to video by adding a transformer decoder. The decoder propagates object representations recurrently through time by using the output tokens for one frame as the object queries for the next. Our model is end-to-end trainable on video data and enjoys improved temporal consistency compared to tracking-by-detection baselines, while retaining the open-world capabilities of the backbone detector. We evaluate our model on the challenging TAO-OW benchmark and demonstrate that open-world capabilities, learned from large-scale image-text pre-training, can be transferred successfully to open-world localization across diverse videos.

The ability to perceive how objects change over time is a crucial ingredient in human intelligence. However, current benchmarks cannot faithfully reflect the temporal understanding abilities of video-language models (VidLMs) due to the existence of static visual shortcuts. To remedy this issue, we present VITATECS, a diagnostic VIdeo-Text dAtaset for the evaluation of TEmporal Concept underStanding. Specifically, we first introduce a fine-grained taxonomy of temporal concepts in natural language in order to diagnose the capability of VidLMs to comprehend different temporal aspects. Furthermore, to disentangle the correlation between static and temporal information, we generate counterfactual video descriptions that differ from the original one only in the specified temporal aspect. We employ a semi-automatic data collection framework using large language models and human-in-the-loop annotation to obtain high-quality counterfactual descriptions efficiently. Evaluation of representative video-language understanding models confirms their deficiency in temporal understanding, revealing the need for greater emphasis on the temporal elements in video-language research.

Learning discriminative spatiotemporal representation is the key problem of video understanding. Recently, Vision Transformers (ViTs) have shown their power in learning long-term video dependency with self-attention. Unfortunately, they exhibit limitations in tackling local video redundancy, due to the blind global comparison among tokens. UniFormer has successfully alleviated this issue, by unifying convolution and self-attention as a relation aggregator in the transformer format. However, this model has to require a tiresome and complicated image-pretraining phrase, before being finetuned on videos. This blocks its wide usage in practice. On the contrary, open-sourced ViTs are readily available and well-pretrained with rich image supervision. Based on these observations, we propose a generic paradigm to build a powerful family of video networks, by arming the pretrained ViTs with efficient UniFormer designs. We call this family UniFormerV2, since it inherits the concise style of the UniFormer block. But it contains brand-new local and global relation aggregators, which allow for preferable accuracy-computation balance by seamlessly integrating advantages from both ViTs and UniFormer. Without any bells and whistles, our UniFormerV2 gets the state-of-the-art recognition performance on 8 popular video benchmarks, including scene-related Kinetics-400/600/700 and Moments in Time, temporal-related Something-Something V1/V2, untrimmed ActivityNet and HACS. In particular, it is the first model to achieve 90% top-1 accuracy on Kinetics-400, to our best knowledge. Code will be available at https://github.com/OpenGVLab/UniFormerV2.

Video transformers have become the de facto standard for human action recognition, yet their exclusive reliance on the RGB modality still limits their adoption in certain domains. One such domain is Activities of Daily Living (ADL), where RGB alone is not sufficient to distinguish between visually similar actions, or actions observed from multiple viewpoints. To facilitate the adoption of video transformers for ADL, we hypothesize that the augmentation of RGB with human pose information, known for its sensitivity to fine-grained motion and multiple viewpoints, is essential. Consequently, we introduce the first Pose Induced Video Transformer: PI-ViT (or pi-ViT), a novel approach that augments the RGB representations learned by video transformers with 2D and 3D pose information. The key elements of pi-ViT are two plug-in modules, 2D Skeleton Induction Module and 3D Skeleton Induction Module, that are responsible for inducing 2D and 3D pose information into the RGB representations. These modules operate by performing pose-aware auxiliary tasks, a design choice that allows pi-ViT to discard the modules during inference. Notably, pi-ViT achieves the state-of-the-art performance on three prominent ADL datasets, encompassing both real-world and large-scale RGB-D datasets, without requiring poses or additional computational overhead at inference.

Learning specific hands-on skills such as cooking, car maintenance, and home repairs increasingly happens via instructional videos. The user experience with such videos is known to be improved by meta-information such as time-stamped annotations for the main steps involved. Generating such annotations automatically is challenging, and we describe here two relevant contributions. First, we construct and release a new dense video captioning dataset, Video Timeline Tags (ViTT), featuring a variety of instructional videos together with time-stamped annotations. Second, we explore several multimodal sequence-to-sequence pretraining strategies that leverage large unsupervised datasets of videos and caption-like texts. We pretrain and subsequently finetune dense video captioning models using both YouCook2 and ViTT. We show that such models generalize well and are robust over a wide variety of instructional videos.

Due to lack of fully publicly available text-to-video models, current video editing methods tend to build on pre-trained text-to-image generation models, however, they still face grand challenges in dealing with the local editing of video with temporal information. First, although existing methods attempt to focus on local area editing by a pre-defined mask, the preservation of the outside-area background is non-ideal due to the spatially entire generation of each frame. In addition, specially providing a mask by user is an additional costly undertaking, so an autonomous masking strategy integrated into the editing process is desirable. Last but not least, image-level pretrained model hasn't learned temporal information across frames of a video which is vital for expressing the motion and dynamics. In this paper, we propose to adapt a image-level blended latent diffusion model to perform local video editing tasks. Specifically, we leverage DDIM inversion to acquire the latents as background latents instead of the randomly noised ones to better preserve the background information of the input video. We further introduce an autonomous mask manufacture mechanism derived from cross-attention maps in diffusion steps. Finally, we enhance the temporal consistency across video frames by transforming the self-attention blocks of U-Net into temporal-spatial blocks. Through extensive experiments, our proposed approach demonstrates effectiveness in different real-world video editing tasks.

Self-supervised learning is a promising paradigm in deep learning that enables learning from unlabeled data by constructing pretext tasks that require learning useful representations. In natural language processing, the dominant pretext task has been masked language modeling (MLM), while in computer vision there exists an equivalent called Masked Image Modeling (MIM). However, MIM is challenging because it requires predicting semantic content in accurate locations. E.g, given an incomplete picture of a dog, we can guess that there is a tail, but we cannot determine its exact location. In this work, we propose FlexPredict, a stochastic model that addresses this challenge by incorporating location uncertainty into the model. Specifically, we condition the model on stochastic masked token positions to guide the model toward learning features that are more robust to location uncertainties. Our approach improves downstream performance on a range of tasks, e.g, compared to MIM baselines, FlexPredict boosts ImageNet linear probing by 1.6% with ViT-B and by 2.5% for semi-supervised video segmentation using ViT-L.

Video try-on stands as a promising area for its tremendous real-world potential. Prior works are limited to transferring product clothing images onto person videos with simple poses and backgrounds, while underperforming on casually captured videos. Recently, Sora revealed the scalability of Diffusion Transformer (DiT) in generating lifelike videos featuring real-world scenarios. Inspired by this, we explore and propose the first DiT-based video try-on framework for practical in-the-wild applications, named VITON-DiT. Specifically, VITON-DiT consists of a garment extractor, a Spatial-Temporal denoising DiT, and an identity preservation ControlNet. To faithfully recover the clothing details, the extracted garment features are fused with the self-attention outputs of the denoising DiT and the ControlNet. We also introduce novel random selection strategies during training and an Interpolated Auto-Regressive (IAR) technique at inference to facilitate long video generation. Unlike existing attempts that require the laborious and restrictive construction of a paired training dataset, severely limiting their scalability, VITON-DiT alleviates this by relying solely on unpaired human dance videos and a carefully designed multi-stage training strategy. Furthermore, we curate a challenging benchmark dataset to evaluate the performance of casual video try-on. Extensive experiments demonstrate the superiority of VITON-DiT in generating spatio-temporal consistent try-on results for in-the-wild videos with complicated human poses.

Machine learning has celebrated a lot of achievements on computer vision tasks such as object detection, but the traditionally used models work with relatively low resolution images. The resolution of recording devices is gradually increasing and there is a rising need for new methods of processing high resolution data. We propose an attention pipeline method which uses two staged evaluation of each image or video frame under rough and refined resolution to limit the total number of necessary evaluations. For both stages, we make use of the fast object detection model YOLO v2. We have implemented our model in code, which distributes the work across GPUs. We maintain high accuracy while reaching the average performance of 3-6 fps on 4K video and 2 fps on 8K video.

We propose a new transformer-based image and video tokenizer with Binary Spherical Quantization (BSQ). BSQ projects the high-dimensional visual embedding to a lower-dimensional hypersphere and then applies binary quantization. BSQ is (1) parameter-efficient without an explicit codebook, (2) scalable to arbitrary token dimensions, and (3) compact: compressing visual data by up to 100times with minimal distortion. Our tokenizer uses a transformer encoder and decoder with simple block-wise causal masking to support variable-length videos as input. The resulting BSQ-ViT achieves state-of-the-art visual reconstruction quality on image and video reconstruction benchmarks with 2.4times throughput compared to the best prior methods. Furthermore, by learning an autoregressive prior for adaptive arithmetic coding, BSQ-ViT achieves comparable results on video compression with state-of-the-art video compression standards. BSQ-ViT also enables masked language models to achieve competitive image synthesis quality to GAN- and diffusion-based methods.

Streaming video clips with large-scale video tokens impede vision transformers (ViTs) for efficient recognition, especially in video action detection where sufficient spatiotemporal representations are required for precise actor identification. In this work, we propose an end-to-end framework for efficient video action detection (EVAD) based on vanilla ViTs. Our EVAD consists of two specialized designs for video action detection. First, we propose a spatiotemporal token dropout from a keyframe-centric perspective. In a video clip, we maintain all tokens from its keyframe, preserve tokens relevant to actor motions from other frames, and drop out the remaining tokens in this clip. Second, we refine scene context by leveraging remaining tokens for better recognizing actor identities. The region of interest (RoI) in our action detector is expanded into temporal domain. The captured spatiotemporal actor identity representations are refined via scene context in a decoder with the attention mechanism. These two designs make our EVAD efficient while maintaining accuracy, which is validated on three benchmark datasets (i.e., AVA, UCF101-24, JHMDB). Compared to the vanilla ViT backbone, our EVAD reduces the overall GFLOPs by 43% and improves real-time inference speed by 40% with no performance degradation. Moreover, even at similar computational costs, our EVAD can improve the performance by 1.1 mAP with higher resolution inputs. Code is available at https://github.com/MCG-NJU/EVAD.

Video virtual try-on aims to generate realistic sequences that maintain garment identity and adapt to a person's pose and body shape in source videos. Traditional image-based methods, relying on warping and blending, struggle with complex human movements and occlusions, limiting their effectiveness in video try-on applications. Moreover, video-based models require extensive, high-quality data and substantial computational resources. To tackle these issues, we reconceptualize video try-on as a process of generating videos conditioned on garment descriptions and human motion. Our solution, WildVidFit, employs image-based controlled diffusion models for a streamlined, one-stage approach. This model, conditioned on specific garments and individuals, is trained on still images rather than videos. It leverages diffusion guidance from pre-trained models including a video masked autoencoder for segment smoothness improvement and a self-supervised model for feature alignment of adjacent frame in the latent space. This integration markedly boosts the model's ability to maintain temporal coherence, enabling more effective video try-on within an image-based framework. Our experiments on the VITON-HD and DressCode datasets, along with tests on the VVT and TikTok datasets, demonstrate WildVidFit's capability to generate fluid and coherent videos. The project page website is at wildvidfit-project.github.io.

Instructional videos provide detailed how-to guides for various tasks, with viewers often posing questions regarding the content. Addressing these questions is vital for comprehending the content, yet receiving immediate answers is difficult. While numerous computational models have been developed for Video Question Answering (Video QA) tasks, they are primarily trained on questions generated based on video content, aiming to produce answers from within the content. However, in real-world situations, users may pose questions that go beyond the video's informational boundaries, highlighting the necessity to determine if a video can provide the answer. Discerning whether a question can be answered by video content is challenging due to the multi-modal nature of videos, where visual and verbal information are intertwined. To bridge this gap, we present the YTCommentQA dataset, which contains naturally-generated questions from YouTube, categorized by their answerability and required modality to answer -- visual, script, or both. Experiments with answerability classification tasks demonstrate the complexity of YTCommentQA and emphasize the need to comprehend the combined role of visual and script information in video reasoning. The dataset is available at https://github.com/lgresearch/YTCommentQA.

The rapid advancement of deep learning models that can generate and synthesis hyper-realistic videos known as Deepfakes and their ease of access to the general public have raised concern from all concerned bodies to their possible malicious intent use. Deep learning techniques can now generate faces, swap faces between two subjects in a video, alter facial expressions, change gender, and alter facial features, to list a few. These powerful video manipulation methods have potential use in many fields. However, they also pose a looming threat to everyone if used for harmful purposes such as identity theft, phishing, and scam. In this work, we propose a Convolutional Vision Transformer for the detection of Deepfakes. The Convolutional Vision Transformer has two components: Convolutional Neural Network (CNN) and Vision Transformer (ViT). The CNN extracts learnable features while the ViT takes in the learned features as input and categorizes them using an attention mechanism. We trained our model on the DeepFake Detection Challenge Dataset (DFDC) and have achieved 91.5 percent accuracy, an AUC value of 0.91, and a loss value of 0.32. Our contribution is that we have added a CNN module to the ViT architecture and have achieved a competitive result on the DFDC dataset.

Talking head synthesis is vital for virtual avatars and human-computer interaction. However, most existing methods are typically limited to accepting control from a single primary modality, restricting their practical utility. To this end, we introduce ACTalker, an end-to-end video diffusion framework that supports both multi-signals control and single-signal control for talking head video generation. For multiple control, we design a parallel mamba structure with multiple branches, each utilizing a separate driving signal to control specific facial regions. A gate mechanism is applied across all branches, providing flexible control over video generation. To ensure natural coordination of the controlled video both temporally and spatially, we employ the mamba structure, which enables driving signals to manipulate feature tokens across both dimensions in each branch. Additionally, we introduce a mask-drop strategy that allows each driving signal to independently control its corresponding facial region within the mamba structure, preventing control conflicts. Experimental results demonstrate that our method produces natural-looking facial videos driven by diverse signals and that the mamba layer seamlessly integrates multiple driving modalities without conflict.

Although action recognition systems can achieve top performance when evaluated on in-distribution test points, they are vulnerable to unanticipated distribution shifts in test data. However, test-time adaptation of video action recognition models against common distribution shifts has so far not been demonstrated. We propose to address this problem with an approach tailored to spatio-temporal models that is capable of adaptation on a single video sample at a step. It consists in a feature distribution alignment technique that aligns online estimates of test set statistics towards the training statistics. We further enforce prediction consistency over temporally augmented views of the same test video sample. Evaluations on three benchmark action recognition datasets show that our proposed technique is architecture-agnostic and able to significantly boost the performance on both, the state of the art convolutional architecture TANet and the Video Swin Transformer. Our proposed method demonstrates a substantial performance gain over existing test-time adaptation approaches in both evaluations of a single distribution shift and the challenging case of random distribution shifts. Code will be available at https://github.com/wlin-at/ViTTA.

Generating fine-grained video descriptions is a fundamental challenge in video understanding. In this work, we introduce Tarsier, a family of large-scale video-language models designed to generate high-quality video descriptions. Tarsier employs CLIP-ViT to encode frames separately and then uses an LLM to model temporal relationships. Despite its simple architecture, we demonstrate that with a meticulously designed two-stage training procedure, the Tarsier models exhibit substantially stronger video description capabilities than any existing open-source model, showing a +51.4% advantage in human side-by-side evaluation over the strongest model. Additionally, they are comparable to state-of-the-art proprietary models, with a +12.3% advantage against GPT-4V and a -6.7% disadvantage against Gemini 1.5 Pro. Besides video description, Tarsier proves to be a versatile generalist model, achieving new state-of-the-art results across nine public benchmarks, including multi-choice VQA, open-ended VQA, and zero-shot video captioning. Our second contribution is the introduction of a new benchmark for evaluating video description models, consisting of a new challenging dataset featuring videos from diverse sources and varying complexity, along with an automatic method specifically designed to assess the quality of fine-grained video descriptions. We make our models and evaluation benchmark publicly available at https://github.com/bytedance/tarsier.

The current popular methods for video object segmentation (VOS) implement feature matching through several hand-crafted modules that separately perform feature extraction and matching. However, the above hand-crafted designs empirically cause insufficient target interaction, thus limiting the dynamic target-aware feature learning in VOS. To tackle these limitations, this paper presents a scalable Simplified VOS (SimVOS) framework to perform joint feature extraction and matching by leveraging a single transformer backbone. Specifically, SimVOS employs a scalable ViT backbone for simultaneous feature extraction and matching between query and reference features. This design enables SimVOS to learn better target-ware features for accurate mask prediction. More importantly, SimVOS could directly apply well-pretrained ViT backbones (e.g., MAE) for VOS, which bridges the gap between VOS and large-scale self-supervised pre-training. To achieve a better performance-speed trade-off, we further explore within-frame attention and propose a new token refinement module to improve the running speed and save computational cost. Experimentally, our SimVOS achieves state-of-the-art results on popular video object segmentation benchmarks, i.e., DAVIS-2017 (88.0% J&F), DAVIS-2016 (92.9% J&F) and YouTube-VOS 2019 (84.2% J&F), without applying any synthetic video or BL30K pre-training used in previous VOS approaches.

Video Temporal Grounding (VTG) is a crucial capability for video understanding models and plays a vital role in downstream tasks such as video browsing and editing. To effectively handle various tasks simultaneously and enable zero-shot prediction, there is a growing trend in employing video LLMs for VTG tasks. However, current video LLM-based methods rely exclusively on natural language generation, lacking the ability to model the clear structure inherent in videos, which restricts their effectiveness in tackling VTG tasks. To address this issue, this paper first formally introduces causal event modeling framework, which represents videos as sequences of events, and predict the current event using previous events, video inputs, and textural instructions. Each event consists of three components: timestamps, salient scores, and textual captions. We then propose a novel task-interleaved video LLM called TRACE to effectively implement the causal event modeling framework in practice. The TRACE processes visual frames, timestamps, salient scores, and text as distinct tasks, employing various encoders and decoding heads for each. Task tokens are arranged in an interleaved sequence according to the causal event modeling framework's formulation. Extensive experiments on various VTG tasks and datasets demonstrate the superior performance of TRACE compared to state-of-the-art video LLMs. Our model and code are available at https://github.com/gyxxyg/TRACE.

Most transformer-based video encoders are limited to short temporal contexts due to their quadratic complexity. While various attempts have been made to extend this context, this has often come at the cost of both conceptual and computational complexity. We propose to instead re-purpose existing pre-trained video transformers by simply fine-tuning them to attend to memories derived non-parametrically from past activations. By leveraging redundancy reduction, our memory-consolidated vision transformer (MC-ViT) effortlessly extends its context far into the past and exhibits excellent scaling behavior when learning from longer videos. In doing so, MC-ViT sets a new state-of-the-art in long-context video understanding on EgoSchema, Perception Test, and Diving48, outperforming methods that benefit from orders of magnitude more parameters.

Video Foundation Models (VFMs) have received limited exploration due to high computational costs and data scarcity. Previous VFMs rely on Image Foundation Models (IFMs), which face challenges in transferring to the video domain. Although VideoMAE has trained a robust ViT from limited data, its low-level reconstruction poses convergence difficulties and conflicts with high-level cross-modal alignment. This paper proposes a training-efficient method for temporal-sensitive VFMs that integrates the benefits of existing methods. To increase data efficiency, we mask out most of the low-semantics video tokens, but selectively align the unmasked tokens with IFM, which serves as the UnMasked Teacher (UMT). By providing semantic guidance, our method enables faster convergence and multimodal friendliness. With a progressive pre-training framework, our model can handle various tasks including scene-related, temporal-related, and complex video-language understanding. Using only public sources for pre-training in 6 days on 32 A100 GPUs, our scratch-built ViT-L/16 achieves state-of-the-art performances on various video tasks. The code and models will be released at https://github.com/OpenGVLab/unmasked_teacher.

Surgical video segmentation is a critical task in computer-assisted surgery and is vital for enhancing surgical quality and patient outcomes. Recently, the Segment Anything Model 2 (SAM2) framework has shown superior advancements in image and video segmentation. However, SAM2 struggles with efficiency due to the high computational demands of processing high-resolution images and complex and long-range temporal dynamics in surgical videos. To address these challenges, we introduce Surgical SAM 2 (SurgSAM-2), an advanced model to utilize SAM2 with an Efficient Frame Pruning (EFP) mechanism, to facilitate real-time surgical video segmentation. The EFP mechanism dynamically manages the memory bank by selectively retaining only the most informative frames, reducing memory usage and computational cost while maintaining high segmentation accuracy. Our extensive experiments demonstrate that SurgSAM-2 significantly improves both efficiency and segmentation accuracy compared to the vanilla SAM2. Remarkably, SurgSAM-2 achieves a 3times FPS compared with SAM2, while also delivering state-of-the-art performance after fine-tuning with lower-resolution data. These advancements establish SurgSAM-2 as a leading model for surgical video analysis, making real-time surgical video segmentation in resource-constrained environments a feasible reality.

Recent developments of vision large language models (LLMs) have seen remarkable progress, yet still encounter challenges towards multimodal generalists, such as coarse-grained instance-level understanding, lack of unified support for both images and videos, and insufficient coverage across various vision tasks. In this paper, we present VITRON, a universal pixel-level vision LLM designed for comprehensive understanding, generating, segmenting, and editing of both static images and dynamic videos. Building on top of an LLM backbone, VITRON incorporates encoders for images, videos, and pixel-level regional visuals within its frontend modules, while employing state-of-the-art visual specialists as its backend, via which VITRON supports a spectrum of vision end tasks, spanning visual comprehension to visual generation, from low level to high level. To ensure an effective and precise message passing from LLM to backend modules for function invocation, we propose a novel hybrid method by simultaneously integrating discrete textual instructions and continuous signal embeddings. Further, we design various pixel-level spatiotemporal vision-language alignment learning for VITRON to reach the best fine-grained visual capability. Finally, a cross-task synergy module is advised to learn to maximize the task-invariant fine-grained visual features, enhancing the synergy between different visual tasks. Demonstrated over 12 visual tasks and evaluated across 22 datasets, VITRON showcases its extensive capabilities in the four main vision task clusters. Overall, this work illuminates the great potential of developing a more unified multimodal generalist. Project homepage: https://vitron-llm.github.io/

Recent vision-language models are driven by large-scale pretrained models. However, adapting pretrained models on limited data presents challenges such as overfitting, catastrophic forgetting, and the cross-modal gap between vision and language. We introduce a parameter-efficient method to address these challenges, combining multimodal prompt learning and a transformer-based mapping network, while keeping the pretrained models frozen. Our experiments on several video question answering benchmarks demonstrate the superiority of our approach in terms of performance and parameter efficiency on both zero-shot and few-shot settings. Our code is available at https://engindeniz.github.io/vitis.

Enabling large language models (LLMs) to read videos is vital for multimodal LLMs. Existing works show promise on short videos whereas long video (longer than e.g.~1 minute) comprehension remains challenging. The major problem lies in the over-compression of videos, i.e., the encoded video representations are not enough to represent the whole video. To address this issue, we propose Long Video Chat (LVChat), where Frame-Scalable Encoding (FSE) is introduced to dynamically adjust the number of embeddings in alignment with the duration of the video to ensure long videos are not overly compressed into a few embeddings. To deal with long videos whose length is beyond videos seen during training, we propose Interleaved Frame Encoding (IFE), repeating positional embedding and interleaving multiple groups of videos to enable long video input, avoiding performance degradation due to overly long videos. Experimental results show that LVChat significantly outperforms existing methods by up to 27\% in accuracy on long-video QA datasets and long-video captioning benchmarks. Our code is published at https://github.com/wangyu-ustc/LVChat.

Recent advances in diffusion models bring new vitality to visual content creation. However, current text-to-video generation models still face significant challenges such as high training costs, substantial data requirements, and difficulties in maintaining consistency between given text and motion of the foreground object. To address these challenges, we propose mask-guided video generation, which can control video generation through mask motion sequences, while requiring limited training data. Our model enhances existing architectures by incorporating foreground masks for precise text-position matching and motion trajectory control. Through mask motion sequences, we guide the video generation process to maintain consistent foreground objects throughout the sequence. Additionally, through a first-frame sharing strategy and autoregressive extension approach, we achieve more stable and longer video generation. Extensive qualitative and quantitative experiments demonstrate that this approach excels in various video generation tasks, such as video editing and generating artistic videos, outperforming previous methods in terms of consistency and quality. Our generated results can be viewed in the supplementary materials.

We introduce Long-VITA, a simple yet effective large multi-modal model for long-context visual-language understanding tasks. It is adept at concurrently processing and analyzing modalities of image, video, and text over 4K frames or 1M tokens while delivering advanced performances on short-context multi-modal tasks. We propose an effective multi-modal training schema that starts with large language models and proceeds through vision-language alignment, general knowledge learning, and two sequential stages of long-sequence fine-tuning. We further implement context-parallelism distributed inference and logits-masked language modeling head to scale Long-VITA to infinitely long inputs of images and texts during model inference. Regarding training data, Long-VITA is built on a mix of 17M samples from public datasets only and demonstrates the state-of-the-art performance on various multi-modal benchmarks, compared against recent cutting-edge models with internal data. Long-VITA is fully reproducible and supports both NPU and GPU platforms for training and testing. By leveraging our inference designs, Long-VITA models achieve a remarkable 2x prefill speedup and 4x context length extension in single node with 8 GPUs. We hope Long-VITA can serve as a competitive baseline and offer valuable insights for the open-source community in advancing long-context multi-modal understanding.

The discrimination of instance embeddings plays a vital role in associating instances across time for online video instance segmentation (VIS). Instance embedding learning is directly supervised by the contrastive loss computed upon the contrastive items (CIs), which are sets of anchor/positive/negative embeddings. Recent online VIS methods leverage CIs sourced from one reference frame only, which we argue is insufficient for learning highly discriminative embeddings. Intuitively, a possible strategy to enhance CIs is replicating the inference phase during training. To this end, we propose a simple yet effective training strategy, called Consistent Training for Online VIS (CTVIS), which devotes to aligning the training and inference pipelines in terms of building CIs. Specifically, CTVIS constructs CIs by referring inference the momentum-averaged embedding and the memory bank storage mechanisms, and adding noise to the relevant embeddings. Such an extension allows a reliable comparison between embeddings of current instances and the stable representations of historical instances, thereby conferring an advantage in modeling VIS challenges such as occlusion, re-identification, and deformation. Empirically, CTVIS outstrips the SOTA VIS models by up to +5.0 points on three VIS benchmarks, including YTVIS19 (55.1% AP), YTVIS21 (50.1% AP) and OVIS (35.5% AP). Furthermore, we find that pseudo-videos transformed from images can train robust models surpassing fully-supervised ones.

Our objective in this work is video-text retrieval - in particular a joint embedding that enables efficient text-to-video retrieval. The challenges in this area include the design of the visual architecture and the nature of the training data, in that the available large scale video-text training datasets, such as HowTo100M, are noisy and hence competitive performance is achieved only at scale through large amounts of compute. We address both these challenges in this paper. We propose an end-to-end trainable model that is designed to take advantage of both large-scale image and video captioning datasets. Our model is an adaptation and extension of the recent ViT and Timesformer architectures, and consists of attention in both space and time. The model is flexible and can be trained on both image and video text datasets, either independently or in conjunction. It is trained with a curriculum learning schedule that begins by treating images as 'frozen' snapshots of video, and then gradually learns to attend to increasing temporal context when trained on video datasets. We also provide a new video-text pretraining dataset WebVid-2M, comprised of over two million videos with weak captions scraped from the internet. Despite training on datasets that are an order of magnitude smaller, we show that this approach yields state-of-the-art results on standard downstream video-retrieval benchmarks including MSR-VTT, MSVD, DiDeMo and LSMDC.

Recent Multimodal Large Language Models (MLLMs) have typically focused on integrating visual and textual modalities, with less emphasis placed on the role of speech in enhancing interaction. However, speech plays a crucial role in multimodal dialogue systems, and implementing high-performance in both vision and speech tasks remains a significant challenge due to the fundamental modality differences. In this paper, we propose a carefully designed multi-stage training methodology that progressively trains LLM to understand both visual and speech information, ultimately enabling fluent vision and speech interaction. Our approach not only preserves strong vision-language capacity, but also enables efficient speech-to-speech dialogue capabilities without separate ASR and TTS modules, significantly accelerating multimodal end-to-end response speed. By comparing our method against state-of-the-art counterparts across benchmarks for image, video, and speech tasks, we demonstrate that our model is equipped with both strong visual and speech capabilities, making near real-time vision and speech interaction.

World models that forecast environmental changes from actions are vital for autonomous driving models with strong generalization. The prevailing driving world model mainly build on video prediction model. Although these models can produce high-fidelity video sequences with advanced diffusion-based generator, they are constrained by their predictive duration and overall generalization capabilities. In this paper, we explore to solve this problem by combining generation loss with MAE-style feature-level context learning. In particular, we instantiate this target with three key design: (1) A more scalable Diffusion Transformer (DiT) structure trained with extra mask construction task. (2) we devise diffusion-related mask tokens to deal with the fuzzy relations between mask reconstruction and generative diffusion process. (3) we extend mask construction task to spatial-temporal domain by utilizing row-wise mask for shifted self-attention rather than masked self-attention in MAE. Then, we adopt a row-wise cross-view module to align with this mask design. Based on above improvement, we propose MaskGWM: a Generalizable driving World Model embodied with Video Mask reconstruction. Our model contains two variants: MaskGWM-long, focusing on long-horizon prediction, and MaskGWM-mview, dedicated to multi-view generation. Comprehensive experiments on standard benchmarks validate the effectiveness of the proposed method, which contain normal validation of Nuscene dataset, long-horizon rollout of OpenDV-2K dataset and zero-shot validation of Waymo dataset. Quantitative metrics on these datasets show our method notably improving state-of-the-art driving world model.

Surgical scene segmentation is critical in computer-assisted surgery and is vital for enhancing surgical quality and patient outcomes. Recently, referring surgical segmentation is emerging, given its advantage of providing surgeons with an interactive experience to segment the target object. However, existing methods are limited by low efficiency and short-term tracking, hindering their applicability in complex real-world surgical scenarios. In this paper, we introduce ReSurgSAM2, a two-stage surgical referring segmentation framework that leverages Segment Anything Model 2 to perform text-referred target detection, followed by tracking with reliable initial frame identification and diversity-driven long-term memory. For the detection stage, we propose a cross-modal spatial-temporal Mamba to generate precise detection and segmentation results. Based on these results, our credible initial frame selection strategy identifies the reliable frame for the subsequent tracking. Upon selecting the initial frame, our method transitions to the tracking stage, where it incorporates a diversity-driven memory mechanism that maintains a credible and diverse memory bank, ensuring consistent long-term tracking. Extensive experiments demonstrate that ReSurgSAM2 achieves substantial improvements in accuracy and efficiency compared to existing methods, operating in real-time at 61.2 FPS. Our code and datasets will be available at https://github.com/jinlab-imvr/ReSurgSAM2.

Video Temporal Grounding (VTG) focuses on accurately identifying event timestamps within a particular video based on a linguistic query, playing a vital role in downstream tasks such as video browsing and editing. While Video Large Language Models (video LLMs) have made significant progress in understanding video content, they often face challenges in accurately pinpointing timestamps within videos, which limits their performance on VTG tasks. Therefore, to improve video LLMs' ability to effectively locate timestamps, we argue that two critical aspects need to be enhanced. First, it is essential to have high-quality instructional tuning datasets that encompass mainstream VTG tasks. Second, directly incorporating timestamp knowledge into video LLMs is crucial, as it enables models to efficiently comprehend timestamp information. To address these needs, we first introduce VTG-IT-120K, a high-quality and comprehensive instruction tuning dataset that covers VTG tasks such as moment retrieval, dense video captioning, video summarization, and video highlight detection. Furthermore, we propose a specially designed video LLM model for VTG tasks, VTG-LLM, which (1) effectively integrates timestamp knowledge into visual tokens; (2) incorporates absolute-time tokens that specifically handle timestamp knowledge, thereby avoiding concept shifts; and (3) introduces a lightweight, high-performance slot-based token compression method to facilitate the sampling of more video frames. Comprehensive experiments showcase the superior performance of VTG-LLM in comparison to other video LLM methods across various VTG tasks. Our code and datasets are available at https://github.com/gyxxyg/VTG-LLM.

The expanse of information available over the internet makes it difficult to identify whether a specific work is a replica or a duplication of a protected work, especially if we talk about visual representations. Strategies are planned to identify the utilization of the copyrighted image in a report. Still, we want to resolve the issue of involving a copyrighted image in a video and a calculation that could recognize the degree of similarity of the copyrighted picture utilized in the video, even for the pieces of the video that are not featured a lot and in the end perform characterization errands on those edges. Machine learning (ML) and artificial intelligence (AI) are vital to address this problem. Numerous associations have been creating different calculations to screen the identification of copyrighted work. This work means concentrating on those calculations, recognizing designs inside the information, and fabricating a more reasonable model for copyrighted image classification and detection. We have used different algorithms like- Image Processing, Convolutional Neural Networks (CNN), Image hashing, etc. Keywords- Copyright, Artificial Intelligence(AI), Copyrighted Image, Convolutional Neural Network(CNN), Image processing, Degree of similarity, Image Hashing.

Robust tooling and publicly available pre-trained models have helped drive recent advances in mechanistic interpretability for language models. However, similar progress in vision mechanistic interpretability has been hindered by the lack of accessible frameworks and pre-trained weights. We present Prisma (Access the codebase here: https://github.com/Prisma-Multimodal/ViT-Prisma), an open-source framework designed to accelerate vision mechanistic interpretability research, providing a unified toolkit for accessing 75+ vision and video transformers; support for sparse autoencoder (SAE), transcoder, and crosscoder training; a suite of 80+ pre-trained SAE weights; activation caching, circuit analysis tools, and visualization tools; and educational resources. Our analysis reveals surprising findings, including that effective vision SAEs can exhibit substantially lower sparsity patterns than language SAEs, and that in some instances, SAE reconstructions can decrease model loss. Prisma enables new research directions for understanding vision model internals while lowering barriers to entry in this emerging field.

This paper explores feature prediction as a stand-alone objective for unsupervised learning from video and introduces V-JEPA, a collection of vision models trained solely using a feature prediction objective, without the use of pretrained image encoders, text, negative examples, reconstruction, or other sources of supervision. The models are trained on 2 million videos collected from public datasets and are evaluated on downstream image and video tasks. Our results show that learning by predicting video features leads to versatile visual representations that perform well on both motion and appearance-based tasks, without adaption of the model's parameters; e.g., using a frozen backbone. Our largest model, a ViT-H/16 trained only on videos, obtains 81.9% on Kinetics-400, 72.2% on Something-Something-v2, and 77.9% on ImageNet1K.

Existing parameter-efficient fine-tuning (PEFT) methods have achieved significant success on vision transformers (ViTs) adaptation by improving parameter efficiency. However, the exploration of enhancing inference efficiency during adaptation remains underexplored. This limits the broader application of pre-trained ViT models, especially when the model is computationally extensive. In this paper, we propose Dynamic Tuning (DyT), a novel approach to improve both parameter and inference efficiency for ViT adaptation. Specifically, besides using the lightweight adapter modules, we propose a token dispatcher to distinguish informative tokens from less important ones, allowing the latter to dynamically skip the original block, thereby reducing the redundant computation during inference. Additionally, we explore multiple design variants to find the best practice of DyT. Finally, inspired by the mixture-of-experts (MoE) mechanism, we introduce an enhanced adapter to further boost the adaptation performance. We validate DyT across various tasks, including image/video recognition and semantic segmentation. For instance, DyT achieves comparable or even superior performance compared to existing PEFT methods while evoking only 71%-85% of their FLOPs on the VTAB-1K benchmark.

The remarkable multimodal capabilities and interactive experience of GPT-4o underscore their necessity in practical applications, yet open-source models rarely excel in both areas. In this paper, we introduce VITA, the first-ever open-source Multimodal Large Language Model (MLLM) adept at simultaneous processing and analysis of Video, Image, Text, and Audio modalities, and meanwhile has an advanced multimodal interactive experience. Starting from Mixtral 8x7B as a language foundation, we expand its Chinese vocabulary followed by bilingual instruction tuning. We further endow the language model with visual and audio capabilities through two-stage multi-task learning of multimodal alignment and instruction tuning. VITA demonstrates robust foundational capabilities of multilingual, vision, and audio understanding, as evidenced by its strong performance across a range of both unimodal and multimodal benchmarks. Beyond foundational capabilities, we have made considerable progress in enhancing the natural multimodal human-computer interaction experience. To the best of our knowledge, we are the first to exploit non-awakening interaction and audio interrupt in MLLM. VITA is the first step for the open-source community to explore the seamless integration of multimodal understanding and interaction. While there is still lots of work to be done on VITA to get close to close-source counterparts, we hope that its role as a pioneer can serve as a cornerstone for subsequent research. Project Page: https://vita-home.github.io.

Vision Transformers (ViTs) have achieved overwhelming success, yet they suffer from vulnerable resolution scalability, i.e., the performance drops drastically when presented with input resolutions that are unseen during training. We introduce, ResFormer, a framework that is built upon the seminal idea of multi-resolution training for improved performance on a wide spectrum of, mostly unseen, testing resolutions. In particular, ResFormer operates on replicated images of different resolutions and enforces a scale consistency loss to engage interactive information across different scales. More importantly, to alternate among varying resolutions effectively, especially novel ones in testing, we propose a global-local positional embedding strategy that changes smoothly conditioned on input sizes. We conduct extensive experiments for image classification on ImageNet. The results provide strong quantitative evidence that ResFormer has promising scaling abilities towards a wide range of resolutions. For instance, ResFormer-B-MR achieves a Top-1 accuracy of 75.86% and 81.72% when evaluated on relatively low and high resolutions respectively (i.e., 96 and 640), which are 48% and 7.49% better than DeiT-B. We also demonstrate, moreover, ResFormer is flexible and can be easily extended to semantic segmentation, object detection and video action recognition. Code is available at https://github.com/ruitian12/resformer.

This paper introduces InternVid, a large-scale video-centric multimodal dataset that enables learning powerful and transferable video-text representations for multimodal understanding and generation. The InternVid dataset contains over 7 million videos lasting nearly 760K hours, yielding 234M video clips accompanied by detailed descriptions of total 4.1B words. Our core contribution is to develop a scalable approach to autonomously build a high-quality video-text dataset with large language models (LLM), thereby showcasing its efficacy in learning video-language representation at scale. Specifically, we utilize a multi-scale approach to generate video-related descriptions. Furthermore, we introduce ViCLIP, a video-text representation learning model based on ViT-L. Learned on InternVid via contrastive learning, this model demonstrates leading zero-shot action recognition and competitive video retrieval performance. Beyond basic video understanding tasks like recognition and retrieval, our dataset and model have broad applications. They are particularly beneficial for generating interleaved video-text data for learning a video-centric dialogue system, advancing video-to-text and text-to-video generation research. These proposed resources provide a tool for researchers and practitioners interested in multimodal video understanding and generation.

We introduce Vidu, a high-performance text-to-video generator that is capable of producing 1080p videos up to 16 seconds in a single generation. Vidu is a diffusion model with U-ViT as its backbone, which unlocks the scalability and the capability for handling long videos. Vidu exhibits strong coherence and dynamism, and is capable of generating both realistic and imaginative videos, as well as understanding some professional photography techniques, on par with Sora -- the most powerful reported text-to-video generator. Finally, we perform initial experiments on other controllable video generation, including canny-to-video generation, video prediction and subject-driven generation, which demonstrate promising results.

We introduce Token Merging (ToMe), a simple method to increase the throughput of existing ViT models without needing to train. ToMe gradually combines similar tokens in a transformer using a general and light-weight matching algorithm that is as fast as pruning while being more accurate. Off-the-shelf, ToMe can 2x the throughput of state-of-the-art ViT-L @ 512 and ViT-H @ 518 models on images and 2.2x the throughput of ViT-L on video with only a 0.2-0.3% accuracy drop in each case. ToMe can also easily be applied during training, improving in practice training speed up to 2x for MAE fine-tuning on video. Training with ToMe further minimizes accuracy drop, leading to 2x the throughput of ViT-B on audio for only a 0.4% mAP drop. Qualitatively, we find that ToMe merges object parts into one token, even over multiple frames of video. Overall, ToMe's accuracy and speed are competitive with state-of-the-art on images, video, and audio.

Pre-training video transformers on extra large-scale datasets is generally required to achieve premier performance on relatively small datasets. In this paper, we show that video masked autoencoders (VideoMAE) are data-efficient learners for self-supervised video pre-training (SSVP). We are inspired by the recent ImageMAE and propose customized video tube masking with an extremely high ratio. This simple design makes video reconstruction a more challenging self-supervision task, thus encouraging extracting more effective video representations during this pre-training process. We obtain three important findings on SSVP: (1) An extremely high proportion of masking ratio (i.e., 90% to 95%) still yields favorable performance of VideoMAE. The temporally redundant video content enables a higher masking ratio than that of images. (2) VideoMAE achieves impressive results on very small datasets (i.e., around 3k-4k videos) without using any extra data. (3) VideoMAE shows that data quality is more important than data quantity for SSVP. Domain shift between pre-training and target datasets is an important issue. Notably, our VideoMAE with the vanilla ViT can achieve 87.4% on Kinetics-400, 75.4% on Something-Something V2, 91.3% on UCF101, and 62.6% on HMDB51, without using any extra data. Code is available at https://github.com/MCG-NJU/VideoMAE.

Existing monocular depth estimation methods have achieved excellent robustness in diverse scenes, but they can only retrieve affine-invariant depth, up to an unknown scale and shift. However, in some video-based scenarios such as video depth estimation and 3D scene reconstruction from a video, the unknown scale and shift residing in per-frame prediction may cause the depth inconsistency. To solve this problem, we propose a locally weighted linear regression method to recover the scale and shift with very sparse anchor points, which ensures the scale consistency along consecutive frames. Extensive experiments show that our method can boost the performance of existing state-of-the-art approaches by 50% at most over several zero-shot benchmarks. Besides, we merge over 6.3 million RGBD images to train strong and robust depth models. Our produced ResNet50-backbone model even outperforms the state-of-the-art DPT ViT-Large model. Combining with geometry-based reconstruction methods, we formulate a new dense 3D scene reconstruction pipeline, which benefits from both the scale consistency of sparse points and the robustness of monocular methods. By performing the simple per-frame prediction over a video, the accurate 3D scene shape can be recovered.

An ideal model for dense video captioning -- predicting captions localized temporally in a video -- should be able to handle long input videos, predict rich, detailed textual descriptions, and be able to produce outputs before processing the entire video. Current state-of-the-art models, however, process a fixed number of downsampled frames, and make a single full prediction after seeing the whole video. We propose a streaming dense video captioning model that consists of two novel components: First, we propose a new memory module, based on clustering incoming tokens, which can handle arbitrarily long videos as the memory is of a fixed size. Second, we develop a streaming decoding algorithm that enables our model to make predictions before the entire video has been processed. Our model achieves this streaming ability, and significantly improves the state-of-the-art on three dense video captioning benchmarks: ActivityNet, YouCook2 and ViTT. Our code is released at https://github.com/google-research/scenic.

Segmenting long videos into chapters enables users to quickly navigate to the information of their interest. This important topic has been understudied due to the lack of publicly released datasets. To address this issue, we present VidChapters-7M, a dataset of 817K user-chaptered videos including 7M chapters in total. VidChapters-7M is automatically created from videos online in a scalable manner by scraping user-annotated chapters and hence without any additional manual annotation. We introduce the following three tasks based on this data. First, the video chapter generation task consists of temporally segmenting the video and generating a chapter title for each segment. To further dissect the problem, we also define two variants of this task: video chapter generation given ground-truth boundaries, which requires generating a chapter title given an annotated video segment, and video chapter grounding, which requires temporally localizing a chapter given its annotated title. We benchmark both simple baselines and state-of-the-art video-language models for these three tasks. We also show that pretraining on VidChapters-7M transfers well to dense video captioning tasks in both zero-shot and finetuning settings, largely improving the state of the art on the YouCook2 and ViTT benchmarks. Finally, our experiments reveal that downstream performance scales well with the size of the pretraining dataset. Our dataset, code, and models are publicly available at https://antoyang.github.io/vidchapters.html.

Effective video tokenization is critical for scaling transformer models for long videos. Current approaches tokenize videos using space-time patches, leading to excessive tokens and computational inefficiencies. The best token reduction strategies degrade performance and barely reduce the number of tokens when the camera moves. We introduce grounded video tokenization, a paradigm that organizes tokens based on panoptic sub-object trajectories rather than fixed patches. Our method aligns with fundamental perceptual principles, ensuring that tokenization reflects scene complexity rather than video duration. We propose TrajViT, a video encoder that extracts object trajectories and converts them into semantically meaningful tokens, significantly reducing redundancy while maintaining temporal coherence. Trained with contrastive learning, TrajViT significantly outperforms space-time ViT (ViT3D) across multiple video understanding benchmarks, e.g., TrajViT outperforms ViT3D by a large margin of 6% top-5 recall in average at video-text retrieval task with 10x token deduction. We also show TrajViT as a stronger model than ViT3D for being the video encoder for modern VideoLLM, obtaining an average of 5.2% performance improvement across 6 VideoQA benchmarks while having 4x faster training time and 18x less inference FLOPs. TrajViT is the first efficient encoder to consistently outperform ViT3D across diverse video analysis tasks, making it a robust and scalable solution.

We present DINO-Tracker -- a new framework for long-term dense tracking in video. The pillar of our approach is combining test-time training on a single video, with the powerful localized semantic features learned by a pre-trained DINO-ViT model. Specifically, our framework simultaneously adopts DINO's features to fit to the motion observations of the test video, while training a tracker that directly leverages the refined features. The entire framework is trained end-to-end using a combination of self-supervised losses, and regularization that allows us to retain and benefit from DINO's semantic prior. Extensive evaluation demonstrates that our method achieves state-of-the-art results on known benchmarks. DINO-tracker significantly outperforms self-supervised methods and is competitive with state-of-the-art supervised trackers, while outperforming them in challenging cases of tracking under long-term occlusions.

Nuanced expressiveness, particularly through fine-grained hand and facial expressions, is pivotal for enhancing the realism and vitality of digital human representations. In this work, we focus on investigating the expressiveness of human avatars when learned from monocular RGB video; a setting that introduces new challenges in capturing and animating fine-grained details. To this end, we introduce EVA, a drivable human model that meticulously sculpts fine details based on 3D Gaussians and SMPL-X, an expressive parametric human model. Focused on enhancing expressiveness, our work makes three key contributions. First, we highlight the critical importance of aligning the SMPL-X model with RGB frames for effective avatar learning. Recognizing the limitations of current SMPL-X prediction methods for in-the-wild videos, we introduce a plug-and-play module that significantly ameliorates misalignment issues. Second, we propose a context-aware adaptive density control strategy, which is adaptively adjusting the gradient thresholds to accommodate the varied granularity across body parts. Last but not least, we develop a feedback mechanism that predicts per-pixel confidence to better guide the learning of 3D Gaussians. Extensive experiments on two benchmarks demonstrate the superiority of our framework both quantitatively and qualitatively, especially on the fine-grained hand and facial details. See the project website at https://evahuman.github.io

In this work, we discuss evaluating video foundation models in a fair and robust manner. Unlike language or image foundation models, many video foundation models are evaluated with differing parameters (such as sampling rate, number of frames, pretraining steps, etc.), making fair and robust comparisons challenging. Therefore, we present a carefully designed evaluation framework for measuring two core capabilities of video comprehension: appearance and motion understanding. Our findings reveal that existing video foundation models, whether text-supervised like UMT or InternVideo2, or self-supervised like V-JEPA, exhibit limitations in at least one of these capabilities. As an alternative, we introduce TWLV-I, a new video foundation model that constructs robust visual representations for both motion- and appearance-based videos. Based on the average top-1 accuracy of linear probing on five action recognition benchmarks, pretrained only on publicly accessible datasets, our model shows a 4.6%p improvement compared to V-JEPA (ViT-L) and a 7.7%p improvement compared to UMT (ViT-L). Even when compared to much larger models, our model demonstrates a 7.2%p improvement compared to DFN (ViT-H), a 2.7%p improvement compared to V-JEPA~(ViT-H) and a 2.8%p improvement compared to InternVideo2 (ViT-g). We provide embedding vectors obtained by TWLV-I from videos of several commonly used video benchmarks, along with evaluation source code that can directly utilize these embeddings. The code is available on "https://github.com/twelvelabs-io/video-embeddings-evaluation-framework".

Recently, large-scale pre-trained language-image models like CLIP have shown extraordinary capabilities for understanding spatial contents, but naively transferring such models to video recognition still suffers from unsatisfactory temporal modeling capabilities. Existing methods insert tunable structures into or in parallel with the pre-trained model, which either requires back-propagation through the whole pre-trained model and is thus resource-demanding, or is limited by the temporal reasoning capability of the pre-trained structure. In this work, we present DiST, which disentangles the learning of spatial and temporal aspects of videos. Specifically, DiST uses a dual-encoder structure, where a pre-trained foundation model acts as the spatial encoder, and a lightweight network is introduced as the temporal encoder. An integration branch is inserted between the encoders to fuse spatio-temporal information. The disentangled spatial and temporal learning in DiST is highly efficient because it avoids the back-propagation of massive pre-trained parameters. Meanwhile, we empirically show that disentangled learning with an extra network for integration benefits both spatial and temporal understanding. Extensive experiments on five benchmarks show that DiST delivers better performance than existing state-of-the-art methods by convincing gaps. When pre-training on the large-scale Kinetics-710, we achieve 89.7% on Kinetics-400 with a frozen ViT-L model, which verifies the scalability of DiST. Codes and models can be found in https://github.com/alibaba-mmai-research/DiST.

In this work, we introduce Vid2Seq, a multi-modal single-stage dense event captioning model pretrained on narrated videos which are readily-available at scale. The Vid2Seq architecture augments a language model with special time tokens, allowing it to seamlessly predict event boundaries and textual descriptions in the same output sequence. Such a unified model requires large-scale training data, which is not available in current annotated datasets. We show that it is possible to leverage unlabeled narrated videos for dense video captioning, by reformulating sentence boundaries of transcribed speech as pseudo event boundaries, and using the transcribed speech sentences as pseudo event captions. The resulting Vid2Seq model pretrained on the YT-Temporal-1B dataset improves the state of the art on a variety of dense video captioning benchmarks including YouCook2, ViTT and ActivityNet Captions. Vid2Seq also generalizes well to the tasks of video paragraph captioning and video clip captioning, and to few-shot settings. Our code is publicly available at https://antoyang.github.io/vid2seq.html.

Most prior art in visual understanding relies solely on analyzing the "what" (e.g., event recognition) and "where" (e.g., event localization), which in some cases, fails to describe correct contextual relationships between events or leads to incorrect underlying visual attention. Part of what defines us as human and fundamentally different from machines is our instinct to seek causality behind any association, say an event Y that happened as a direct result of event X. To this end, we propose iPerceive, a framework capable of understanding the "why" between events in a video by building a common-sense knowledge base using contextual cues to infer causal relationships between objects in the video. We demonstrate the effectiveness of our technique using the dense video captioning (DVC) and video question answering (VideoQA) tasks. Furthermore, while most prior work in DVC and VideoQA relies solely on visual information, other modalities such as audio and speech are vital for a human observer's perception of an environment. We formulate DVC and VideoQA tasks as machine translation problems that utilize multiple modalities. By evaluating the performance of iPerceive DVC and iPerceive VideoQA on the ActivityNet Captions and TVQA datasets respectively, we show that our approach furthers the state-of-the-art. Code and samples are available at: iperceive.amanchadha.com.

The parameter-efficient adaptation of the image-text pretraining model CLIP for video-text retrieval is a prominent area of research. While CLIP is focused on image-level vision-language matching, video-text retrieval demands comprehensive understanding at the video level. Three key discrepancies emerge in the transfer from image-level to video-level: vision, language, and alignment. However, existing methods mainly focus on vision while neglecting language and alignment. In this paper, we propose Discrepancy Reduction in Vision, Language, and Alignment (DiscoVLA), which simultaneously mitigates all three discrepancies. Specifically, we introduce Image-Video Features Fusion to integrate image-level and video-level features, effectively tackling both vision and language discrepancies. Additionally, we generate pseudo image captions to learn fine-grained image-level alignment. To mitigate alignment discrepancies, we propose Image-to-Video Alignment Distillation, which leverages image-level alignment knowledge to enhance video-level alignment. Extensive experiments demonstrate the superiority of our DiscoVLA. In particular, on MSRVTT with CLIP (ViT-B/16), DiscoVLA outperforms previous methods by 1.5% in R@1, reaching a final score of 50.5% R@1. The code is available at https://github.com/LunarShen/DsicoVLA.

People perceive the world with multiple senses (e.g., through hearing sounds, reading words and seeing objects). However, most existing AI systems only process an individual modality. This paper presents an approach that excels at handling multiple modalities of information with a single model. In our "{SkillNet}" model, different parts of the parameters are specialized for processing different modalities. Unlike traditional dense models that always activate all the model parameters, our model sparsely activates parts of the parameters whose skills are relevant to the task. Such model design enables SkillNet to learn skills in a more interpretable way. We develop our model for five modalities including text, image, sound, video and code. Results show that, SkillNet performs comparably to five modality-specific fine-tuned models. Moreover, our model supports self-supervised pretraining with the same sparsely activated way, resulting in better initialized parameters for different modalities. We find that pretraining significantly improves the performance of SkillNet on five modalities, on par with or even better than baselines with modality-specific pretraining. On the task of Chinese text-to-image retrieval, our final system achieves higher accuracy than existing leading systems including Wukong{ViT-B} and Wenlan 2.0 while using less number of activated parameters.

Multimodal Large Language Models (MLLMs) increasingly excel at perception, understanding, and reasoning. However, current benchmarks inadequately evaluate their ability to perform these tasks continuously in dynamic, real-world environments. To bridge this gap, we introduce RTV-Bench, a fine-grained benchmark for MLLM real-time video analysis. RTV-Bench uses three key principles: (1) Multi-Timestamp Question Answering (MTQA), where answers evolve with scene changes; (2) Hierarchical Question Structure, combining basic and advanced queries; and (3) Multi-dimensional Evaluation, assessing the ability of continuous perception, understanding, and reasoning. RTV-Bench contains 552 diverse videos (167.2 hours) and 4,631 high-quality QA pairs. We evaluated leading MLLMs, including proprietary (GPT-4o, Gemini 2.0), open-source offline (Qwen2.5-VL, VideoLLaMA3), and open-source real-time (VITA-1.5, InternLM-XComposer2.5-OmniLive) models. Experiment results show open-source real-time models largely outperform offline ones but still trail top proprietary models. Our analysis also reveals that larger model size or higher frame sampling rates do not significantly boost RTV-Bench performance, sometimes causing slight decreases. This underscores the need for better model architectures optimized for video stream processing and long sequences to advance real-time video analysis with MLLMs. Our benchmark toolkit is available at: https://github.com/LJungang/RTV-Bench.

Current methods for 3D scene reconstruction from sparse posed images employ intermediate 3D representations such as neural fields, voxel grids, or 3D Gaussians, to achieve multi-view consistent scene appearance and geometry. In this paper we introduce MVGD, a diffusion-based architecture capable of direct pixel-level generation of images and depth maps from novel viewpoints, given an arbitrary number of input views. Our method uses raymap conditioning to both augment visual features with spatial information from different viewpoints, as well as to guide the generation of images and depth maps from novel views. A key aspect of our approach is the multi-task generation of images and depth maps, using learnable task embeddings to guide the diffusion process towards specific modalities. We train this model on a collection of more than 60 million multi-view samples from publicly available datasets, and propose techniques to enable efficient and consistent learning in such diverse conditions. We also propose a novel strategy that enables the efficient training of larger models by incrementally fine-tuning smaller ones, with promising scaling behavior. Through extensive experiments, we report state-of-the-art results in multiple novel view synthesis benchmarks, as well as multi-view stereo and video depth estimation.

Generating video stories from text prompts is a complex task. In addition to having high visual quality, videos need to realistically adhere to a sequence of text prompts whilst being consistent throughout the frames. Creating a benchmark for video generation requires data annotated over time, which contrasts with the single caption used often in video datasets. To fill this gap, we collect comprehensive human annotations on three existing datasets, and introduce StoryBench: a new, challenging multi-task benchmark to reliably evaluate forthcoming text-to-video models. Our benchmark includes three video generation tasks of increasing difficulty: action execution, where the next action must be generated starting from a conditioning video; story continuation, where a sequence of actions must be executed starting from a conditioning video; and story generation, where a video must be generated from only text prompts. We evaluate small yet strong text-to-video baselines, and show the benefits of training on story-like data algorithmically generated from existing video captions. Finally, we establish guidelines for human evaluation of video stories, and reaffirm the need of better automatic metrics for video generation. StoryBench aims at encouraging future research efforts in this exciting new area.

Understanding the physical world is a fundamental challenge in embodied AI, critical for enabling agents to perform complex tasks and operate safely in real-world environments. While Vision-Language Models (VLMs) have shown great promise in reasoning and task planning for embodied agents, their ability to comprehend physical phenomena remains extremely limited. To close this gap, we introduce PhysBench, a comprehensive benchmark designed to evaluate VLMs' physical world understanding capability across a diverse set of tasks. PhysBench contains 10,002 entries of interleaved video-image-text data, categorized into four major domains: physical object properties, physical object relationships, physical scene understanding, and physics-based dynamics, further divided into 19 subclasses and 8 distinct capability dimensions. Our extensive experiments, conducted on 75 representative VLMs, reveal that while these models excel in common-sense reasoning, they struggle with understanding the physical world -- likely due to the absence of physical knowledge in their training data and the lack of embedded physical priors. To tackle the shortfall, we introduce PhysAgent, a novel framework that combines the generalization strengths of VLMs with the specialized expertise of vision models, significantly enhancing VLMs' physical understanding across a variety of tasks, including an 18.4\% improvement on GPT-4o. Furthermore, our results demonstrate that enhancing VLMs' physical world understanding capabilities can help embodied agents such as MOKA. We believe that PhysBench and PhysAgent offer valuable insights and contribute to bridging the gap between VLMs and physical world understanding.

We release a realistic, diverse, and challenging dataset for object detection on images. The data was recorded at a beer tent in Germany and consists of 15 different categories of food and drink items. We created more than 2,500 object annotations by hand for 1,110 images captured by a video camera above the checkout. We further make available the remaining 600GB of (unlabeled) data containing days of footage. Additionally, we provide our trained models as a benchmark. Possible applications include automated checkout systems which could significantly speed up the process.

Short-term action anticipation (STA) in first-person videos is a challenging task that involves understanding the next active object interactions and predicting future actions. Existing action anticipation methods have primarily focused on utilizing features extracted from video clips, but often overlooked the importance of objects and their interactions. To this end, we propose a novel approach that applies a guided attention mechanism between the objects, and the spatiotemporal features extracted from video clips, enhancing the motion and contextual information, and further decoding the object-centric and motion-centric information to address the problem of STA in egocentric videos. Our method, GANO (Guided Attention for Next active Objects) is a multi-modal, end-to-end, single transformer-based network. The experimental results performed on the largest egocentric dataset demonstrate that GANO outperforms the existing state-of-the-art methods for the prediction of the next active object label, its bounding box location, the corresponding future action, and the time to contact the object. The ablation study shows the positive contribution of the guided attention mechanism compared to other fusion methods. Moreover, it is possible to improve the next active object location and class label prediction results of GANO by just appending the learnable object tokens with the region of interest embeddings.

We present Amazon Nova, a new generation of state-of-the-art foundation models that deliver frontier intelligence and industry-leading price performance. Amazon Nova Pro is a highly-capable multimodal model with the best combination of accuracy, speed, and cost for a wide range of tasks. Amazon Nova Lite is a low-cost multimodal model that is lightning fast for processing images, video, documents and text. Amazon Nova Micro is a text-only model that delivers our lowest-latency responses at very low cost. Amazon Nova Canvas is an image generation model that creates professional grade images with rich customization controls. Amazon Nova Reel is a video generation model offering high-quality outputs, customization, and motion control. Our models were built responsibly and with a commitment to customer trust, security, and reliability. We report benchmarking results for core capabilities, agentic performance, long context, functional adaptation, runtime performance, and human evaluation.

In this technical report, we describe the Guided-Attention mechanism based solution for the short-term anticipation (STA) challenge for the EGO4D challenge. It combines the object detections, and the spatiotemporal features extracted from video clips, enhancing the motion and contextual information, and further decoding the object-centric and motion-centric information to address the problem of STA in egocentric videos. For the challenge, we build our model on top of StillFast with Guided Attention applied on fast network. Our model obtains better performance on the validation set and also achieves state-of-the-art (SOTA) results on the challenge test set for EGO4D Short-Term Object Interaction Anticipation Challenge.

Most model-free visual object tracking methods formulate the tracking task as object location estimation given by a 2D segmentation or a bounding box in each video frame. We argue that this representation is limited and instead propose to guide and improve 2D tracking with an explicit object representation, namely the textured 3D shape and 6DoF pose in each video frame. Our representation tackles a complex long-term dense correspondence problem between all 3D points on the object for all video frames, including frames where some points are invisible. To achieve that, the estimation is driven by re-rendering the input video frames as well as possible through differentiable rendering, which has not been used for tracking before. The proposed optimization minimizes a novel loss function to estimate the best 3D shape, texture, and 6DoF pose. We improve the state-of-the-art in 2D segmentation tracking on three different datasets with mostly rigid objects.

Scalable learning of humanoid robots is crucial for their deployment in real-world applications. While traditional approaches primarily rely on reinforcement learning or teleoperation to achieve whole-body control, they are often limited by the diversity of simulated environments and the high costs of demonstration collection. In contrast, human videos are ubiquitous and present an untapped source of semantic and motion information that could significantly enhance the generalization capabilities of humanoid robots. This paper introduces Humanoid-X, a large-scale dataset of over 20 million humanoid robot poses with corresponding text-based motion descriptions, designed to leverage this abundant data. Humanoid-X is curated through a comprehensive pipeline: data mining from the Internet, video caption generation, motion retargeting of humans to humanoid robots, and policy learning for real-world deployment. With Humanoid-X, we further train a large humanoid model, UH-1, which takes text instructions as input and outputs corresponding actions to control a humanoid robot. Extensive simulated and real-world experiments validate that our scalable training approach leads to superior generalization in text-based humanoid control, marking a significant step toward adaptable, real-world-ready humanoid robots.

In this report, we present the latest model of the Gemini family, Gemini 1.5 Pro, a highly compute-efficient multimodal mixture-of-experts model capable of recalling and reasoning over fine-grained information from millions of tokens of context, including multiple long documents and hours of video and audio. Gemini 1.5 Pro achieves near-perfect recall on long-context retrieval tasks across modalities, improves the state-of-the-art in long-document QA, long-video QA and long-context ASR, and matches or surpasses Gemini 1.0 Ultra's state-of-the-art performance across a broad set of benchmarks. Studying the limits of Gemini 1.5 Pro's long-context ability, we find continued improvement in next-token prediction and near-perfect retrieval (>99%) up to at least 10M tokens, a generational leap over existing models such as Claude 2.1 (200k) and GPT-4 Turbo (128k). Finally, we highlight surprising new capabilities of large language models at the frontier; when given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person who learned from the same content.

In this report, we introduce the Gemini 2.X model family: Gemini 2.5 Pro and Gemini 2.5 Flash, as well as our earlier Gemini 2.0 Flash and Flash-Lite models. Gemini 2.5 Pro is our most capable model yet, achieving SoTA performance on frontier coding and reasoning benchmarks. In addition to its incredible coding and reasoning skills, Gemini 2.5 Pro is a thinking model that excels at multimodal understanding and it is now able to process up to 3 hours of video content. Its unique combination of long context, multimodal and reasoning capabilities can be combined to unlock new agentic workflows. Gemini 2.5 Flash provides excellent reasoning abilities at a fraction of the compute and latency requirements and Gemini 2.0 Flash and Flash-Lite provide high performance at low latency and cost. Taken together, the Gemini 2.X model generation spans the full Pareto frontier of model capability vs cost, allowing users to explore the boundaries of what is possible with complex agentic problem solving.

This report introduces a new family of multimodal models, Gemini, that exhibit remarkable capabilities across image, audio, video, and text understanding. The Gemini family consists of Ultra, Pro, and Nano sizes, suitable for applications ranging from complex reasoning tasks to on-device memory-constrained use-cases. Evaluation on a broad range of benchmarks shows that our most-capable Gemini Ultra model advances the state of the art in 30 of 32 of these benchmarks - notably being the first model to achieve human-expert performance on the well-studied exam benchmark MMLU, and improving the state of the art in every one of the 20 multimodal benchmarks we examined. We believe that the new capabilities of Gemini models in cross-modal reasoning and language understanding will enable a wide variety of use cases and we discuss our approach toward deploying them responsibly to users.

Video-based spatial cognition is vital for robotics and embodied AI but challenges current Vision-Language Models (VLMs). This paper makes two key contributions. First, we introduce ViCA (Visuospatial Cognitive Assistant)-322K, a diverse dataset of 322,003 QA pairs from real-world indoor videos (ARKitScenes, ScanNet, ScanNet++), offering supervision for 3D metadata-grounded queries and video-based complex reasoning. Second, we develop ViCA-7B, fine-tuned on ViCA-322K, which achieves new state-of-the-art on all eight VSI-Bench tasks, outperforming existing models, including larger ones (e.g., +26.1 on Absolute Distance). For interpretability, we present ViCA-Thinking-2.68K, a dataset with explicit reasoning chains, and fine-tune ViCA-7B to create ViCA-7B-Thinking, a model that articulates its spatial reasoning. Our work highlights the importance of targeted data and suggests paths for improved temporal-spatial modeling. We release all resources to foster research in robust visuospatial intelligence.

The scaling of Transformers has driven breakthrough capabilities for language models. At present, the largest large language models (LLMs) contain upwards of 100B parameters. Vision Transformers (ViT) have introduced the same architecture to image and video modelling, but these have not yet been successfully scaled to nearly the same degree; the largest dense ViT contains 4B parameters (Chen et al., 2022). We present a recipe for highly efficient and stable training of a 22B-parameter ViT (ViT-22B) and perform a wide variety of experiments on the resulting model. When evaluated on downstream tasks (often with a lightweight linear model on frozen features), ViT-22B demonstrates increasing performance with scale. We further observe other interesting benefits of scale, including an improved tradeoff between fairness and performance, state-of-the-art alignment to human visual perception in terms of shape/texture bias, and improved robustness. ViT-22B demonstrates the potential for "LLM-like" scaling in vision, and provides key steps towards getting there.

Visual tokenization via auto-encoding empowers state-of-the-art image and video generative models by compressing pixels into a latent space. Although scaling Transformer-based generators has been central to recent advances, the tokenizer component itself is rarely scaled, leaving open questions about how auto-encoder design choices influence both its objective of reconstruction and downstream generative performance. Our work aims to conduct an exploration of scaling in auto-encoders to fill in this blank. To facilitate this exploration, we replace the typical convolutional backbone with an enhanced Vision Transformer architecture for Tokenization (ViTok). We train ViTok on large-scale image and video datasets far exceeding ImageNet-1K, removing data constraints on tokenizer scaling. We first study how scaling the auto-encoder bottleneck affects both reconstruction and generation -- and find that while it is highly correlated with reconstruction, its relationship with generation is more complex. We next explored the effect of separately scaling the auto-encoders' encoder and decoder on reconstruction and generation performance. Crucially, we find that scaling the encoder yields minimal gains for either reconstruction or generation, while scaling the decoder boosts reconstruction but the benefits for generation are mixed. Building on our exploration, we design ViTok as a lightweight auto-encoder that achieves competitive performance with state-of-the-art auto-encoders on ImageNet-1K and COCO reconstruction tasks (256p and 512p) while outperforming existing auto-encoders on 16-frame 128p video reconstruction for UCF-101, all with 2-5x fewer FLOPs. When integrated with Diffusion Transformers, ViTok demonstrates competitive performance on image generation for ImageNet-1K and sets new state-of-the-art benchmarks for class-conditional video generation on UCF-101.

Capturing complex temporal relationships between video and audio modalities is vital for Audio-Visual Emotion Recognition (AVER). However, existing methods lack attention to local details, such as facial state changes between video frames, which can reduce the discriminability of features and thus lower recognition accuracy. In this paper, we propose a Detail-Enhanced Intra- and Inter-modal Interaction network(DE-III) for AVER, incorporating several novel aspects. We introduce optical flow information to enrich video representations with texture details that better capture facial state changes. A fusion module integrates the optical flow estimation with the corresponding video frames to enhance the representation of facial texture variations. We also design attentive intra- and inter-modal feature enhancement modules to further improve the richness and discriminability of video and audio representations. A detailed quantitative evaluation shows that our proposed model outperforms all existing methods on three benchmark datasets for both concrete and continuous emotion recognition. To encourage further research and ensure replicability, we will release our full code upon acceptance.

Segment Anything Model 2 (SAM 2) has emerged as a powerful tool for video object segmentation and tracking anything. Key components of SAM 2 that drive the impressive video object segmentation performance include a large multistage image encoder for frame feature extraction and a memory mechanism that stores memory contexts from past frames to help current frame segmentation. The high computation complexity of multistage image encoder and memory module has limited its applications in real-world tasks, e.g., video object segmentation on mobile devices. To address this limitation, we propose EfficientTAMs, lightweight track anything models that produce high-quality results with low latency and model size. Our idea is based on revisiting the plain, nonhierarchical Vision Transformer (ViT) as an image encoder for video object segmentation, and introducing an efficient memory module, which reduces the complexity for both frame feature extraction and memory computation for current frame segmentation. We take vanilla lightweight ViTs and efficient memory module to build EfficientTAMs, and train the models on SA-1B and SA-V datasets for video object segmentation and track anything tasks. We evaluate on multiple video segmentation benchmarks including semi-supervised VOS and promptable video segmentation, and find that our proposed EfficientTAM with vanilla ViT perform comparably to SAM 2 model (HieraB+SAM 2) with ~2x speedup on A100 and ~2.4x parameter reduction. On segment anything image tasks, our EfficientTAMs also perform favorably over original SAM with ~20x speedup on A100 and ~20x parameter reduction. On mobile devices such as iPhone 15 Pro Max, our EfficientTAMs can run at ~10 FPS for performing video object segmentation with reasonable quality, highlighting the capability of small models for on-device video object segmentation applications.

This paper asks whether current self-supervised learning methods, if sufficiently scaled up, would be able to reach human-level visual object recognition capabilities with the same type and amount of visual experience humans learn from. Previous work on this question only considered the scaling of data size. Here, we consider the simultaneous scaling of data size, model size, and image resolution. We perform a scaling experiment with vision transformers up to 633M parameters in size (ViT-H/14) trained with up to 5K hours of human-like video data (long, continuous, mostly egocentric videos) with image resolutions of up to 476x476 pixels. The efficiency of masked autoencoders (MAEs) as a self-supervised learning algorithm makes it possible to run this scaling experiment on an unassuming academic budget. We find that it is feasible to reach human-level object recognition capacity at sub-human scales of model size, data size, and image size, if these factors are scaled up simultaneously. To give a concrete example, we estimate that a 2.5B parameter ViT model trained with 20K hours (2.3 years) of human-like video data with a spatial resolution of 952x952 pixels should be able to reach roughly human-level accuracy on ImageNet. Human-level competence is thus achievable for a fundamental perceptual capability from human-like perceptual experience (human-like in both amount and type) with extremely generic learning algorithms and architectures and without any substantive inductive biases.

Transformers were originally proposed as a sequence-to-sequence model for text but have become vital for a wide range of modalities, including images, audio, video, and undirected graphs. However, transformers for directed graphs are a surprisingly underexplored topic, despite their applicability to ubiquitous domains including source code and logic circuits. In this work, we propose two direction- and structure-aware positional encodings for directed graphs: (1) the eigenvectors of the Magnetic Laplacian - a direction-aware generalization of the combinatorial Laplacian; (2) directional random walk encodings. Empirically, we show that the extra directionality information is useful in various downstream tasks, including correctness testing of sorting networks and source code understanding. Together with a data-flow-centric graph construction, our model outperforms the prior state of the art on the Open Graph Benchmark Code2 relatively by 14.7%.

In the field of multi-modal language models, the majority of methods are built on an architecture similar to LLaVA. These models use a single-layer ViT feature as a visual prompt, directly feeding it into the language models alongside textual tokens. However, when dealing with long sequences of visual signals or inputs such as videos, the self-attention mechanism of language models can lead to significant computational overhead. Additionally, using single-layer ViT features makes it challenging for large language models to perceive visual signals fully. This paper proposes an efficient multi-modal language model to minimize computational costs while enabling the model to perceive visual signals as comprehensively as possible. Our method primarily includes: (1) employing cross-attention to image-text interaction similar to Flamingo. (2) utilize hierarchical ViT features. (3) introduce the Mixture of Experts (MoE) mechanism to enhance model effectiveness. Our model achieves competitive scores on public multi-modal benchmarks and performs well in tasks such as image captioning and video captioning.

Despite the remarkable success of Vision Transformers (ViTs) in various visual tasks, they are often hindered by substantial computational cost. In this work, we introduce Vote\&Mix (VoMix), a plug-and-play and parameter-free token reduction method, which can be readily applied to off-the-shelf ViT models without any training. VoMix tackles the computational redundancy of ViTs by identifying tokens with high homogeneity through a layer-wise token similarity voting mechanism. Subsequently, the selected tokens are mixed into the retained set, thereby preserving visual information. Experiments demonstrate VoMix significantly improves the speed-accuracy tradeoff of ViTs on both images and videos. Without any training, VoMix achieves a 2times increase in throughput of existing ViT-H on ImageNet-1K and a 2.4times increase in throughput of existing ViT-L on Kinetics-400 video dataset, with a mere 0.3\% drop in top-1 accuracy.

The current landscape of research leveraging large language models (LLMs) is experiencing a surge. Many works harness the powerful reasoning capabilities of these models to comprehend various modalities, such as text, speech, images, videos, etc. They also utilize LLMs to understand human intention and generate desired outputs like images, videos, and music. However, research that combines both understanding and generation using LLMs is still limited and in its nascent stage. To address this gap, we introduce a Multi-modal Music Understanding and Generation (M^{2}UGen) framework that integrates LLM's abilities to comprehend and generate music for different modalities. The M^{2}UGen framework is purpose-built to unlock creative potential from diverse sources of inspiration, encompassing music, image, and video through the use of pretrained MERT, ViT, and ViViT models, respectively. To enable music generation, we explore the use of AudioLDM 2 and MusicGen. Bridging multi-modal understanding and music generation is accomplished through the integration of the LLaMA 2 model. Furthermore, we make use of the MU-LLaMA model to generate extensive datasets that support text/image/video-to-music generation, facilitating the training of our M^{2}UGen framework. We conduct a thorough evaluation of our proposed framework. The experimental results demonstrate that our model achieves or surpasses the performance of the current state-of-the-art models.

Conventional Vision Transformer simplifies visual modeling by standardizing input resolutions, often disregarding the variability of natural visual data and compromising spatial-contextual fidelity. While preliminary explorations have superficially investigated native resolution modeling, existing approaches still lack systematic analysis from a visual representation perspective. To bridge this gap, we introduce UniViTAR, a family of homogeneous vision foundation models tailored for unified visual modality and native resolution scenario in the era of multimodal. Our framework first conducts architectural upgrades to the vanilla paradigm by integrating multiple advanced components. Building upon these improvements, a progressive training paradigm is introduced, which strategically combines two core mechanisms: (1) resolution curriculum learning, transitioning from fixed-resolution pretraining to native resolution tuning, thereby leveraging ViT's inherent adaptability to variable-length sequences, and (2) visual modality adaptation via inter-batch image-video switching, which balances computational efficiency with enhanced temporal reasoning. In parallel, a hybrid training framework further synergizes sigmoid-based contrastive loss with feature distillation from a frozen teacher model, thereby accelerating early-stage convergence. Finally, trained exclusively on public datasets, externsive experiments across multiple model scales from 0.3B to 1B demonstrate its effectiveness.

We collected a large dataset consisting of nearly 900 unique participants. For every participant we recorded two 30 second uncompressed videos, synchronized PPG waveforms and a single blood pressure measurement. Gender, age and skin color were also registered for every participant. The dataset includes roughly equal numbers of males and females, as well as participants of all ages. While the skin color distribution could have been more balanced, the dataset contains individuals from every skin color. The data was collected in a diverse set of locations to ensure a wide variety of backgrounds and lighting conditions. In an effort to assist in the research and development of remote vital sign measurement we are now opening up access to this dataset.

Causality knowledge is vital to building robust AI systems. Deep learning models often perform poorly on tasks that require causal reasoning, which is often derived using some form of commonsense knowledge not immediately available in the input but implicitly inferred by humans. Prior work has unraveled spurious observational biases that models fall prey to in the absence of causality. While language representation models preserve contextual knowledge within learned embeddings, they do not factor in causal relationships during training. By blending causal relationships with the input features to an existing model that performs visual cognition tasks (such as scene understanding, video captioning, video question-answering, etc.), better performance can be achieved owing to the insight causal relationships bring about. Recently, several models have been proposed that have tackled the task of mining causal data from either the visual or textual modality. However, there does not exist widespread research that mines causal relationships by juxtaposing the visual and language modalities. While images offer a rich and easy-to-process resource for us to mine causality knowledge from, videos are denser and consist of naturally time-ordered events. Also, textual information offers details that could be implicit in videos. We propose iReason, a framework that infers visual-semantic commonsense knowledge using both videos and natural language captions. Furthermore, iReason's architecture integrates a causal rationalization module to aid the process of interpretability, error analysis and bias detection. We demonstrate the effectiveness of iReason using a two-pronged comparative analysis with language representation learning models (BERT, GPT-2) as well as current state-of-the-art multimodal causality models.

Multimodal counterfactual reasoning is a vital yet challenging ability for AI systems. It involves predicting the outcomes of hypothetical circumstances based on vision and language inputs, which enables AI models to learn from failures and explore hypothetical scenarios. Despite its importance, there are only a few datasets targeting the counterfactual reasoning abilities of multimodal models. Among them, they only cover reasoning over synthetic environments or specific types of events (e.g. traffic collisions), making them hard to reliably benchmark the model generalization ability in diverse real-world scenarios and reasoning dimensions. To overcome these limitations, we develop a video question answering dataset, ACQUIRED: it consists of 3.9K annotated videos, encompassing a wide range of event types and incorporating both first and third-person viewpoints, which ensures a focus on real-world diversity. In addition, each video is annotated with questions that span three distinct dimensions of reasoning, including physical, social, and temporal, which can comprehensively evaluate the model counterfactual abilities along multiple aspects. We benchmark our dataset against several state-of-the-art language-only and multimodal models and experimental results demonstrate a significant performance gap (>13%) between models and humans. The findings suggest that multimodal counterfactual reasoning remains an open challenge and ACQUIRED is a comprehensive and reliable benchmark for inspiring future research in this direction.

Diffusion models have been widely used for conditional data cross-modal generation tasks such as text-to-image and text-to-video. However, state-of-the-art models still fail to align the generated visual concepts with high-level semantics in a language such as object count, spatial relationship, etc. We approach this problem from a multimodal data fusion perspective and investigate how different fusion strategies can affect vision-language alignment. We discover that compared to the widely used early fusion of conditioning text in a pretrained image feature space, a specially designed intermediate fusion can: (i) boost text-to-image alignment with improved generation quality and (ii) improve training and inference efficiency by reducing low-rank text-to-image attention calculations. We perform experiments using a text-to-image generation task on the MS-COCO dataset. We compare our intermediate fusion mechanism with the classic early fusion mechanism on two common conditioning methods on a U-shaped ViT backbone. Our intermediate fusion model achieves a higher CLIP Score and lower FID, with 20% reduced FLOPs, and 50% increased training speed compared to a strong U-ViT baseline with an early fusion.

The ability to forecast human-environment collisions from egocentric observations is vital to enable collision avoidance in applications such as VR, AR, and wearable assistive robotics. In this work, we introduce the challenging problem of predicting collisions in diverse environments from multi-view egocentric videos captured from body-mounted cameras. Solving this problem requires a generalizable perception system that can classify which human body joints will collide and estimate a collision region heatmap to localize collisions in the environment. To achieve this, we propose a transformer-based model called COPILOT to perform collision prediction and localization simultaneously, which accumulates information across multi-view inputs through a novel 4D space-time-viewpoint attention mechanism. To train our model and enable future research on this task, we develop a synthetic data generation framework that produces egocentric videos of virtual humans moving and colliding within diverse 3D environments. This framework is then used to establish a large-scale dataset consisting of 8.6M egocentric RGBD frames. Extensive experiments show that COPILOT generalizes to unseen synthetic as well as real-world scenes. We further demonstrate COPILOT outputs are useful for downstream collision avoidance through simple closed-loop control. Please visit our project webpage at https://sites.google.com/stanford.edu/copilot.

The increasing variety and quantity of tagged multimedia content on a variety of online platforms offer a unique opportunity to advance the field of human action recognition. In this study, we utilize 283,582 unique, unlabeled TikTok video clips, categorized into 386 hashtags, to train a domain-specific foundation model for action recognition. We employ VideoMAE V2, an advanced model integrating Masked Autoencoders (MAE) with Vision Transformers (ViT), pre-trained on this diverse collection of unstructured videos. Our model, fine-tuned on established action recognition benchmarks such as UCF101 and HMDB51, achieves state-of-the-art results: 99.05% on UCF101, 86.08% on HMDB51, 85.51% on Kinetics-400, and 74.27% on Something-Something V2 using the ViT-giant backbone. These results highlight the potential of using unstructured and unlabeled videos as a valuable source of diverse and dynamic content for training foundation models. Our investigation confirms that while initial increases in pre-training data volume significantly enhance model performance, the gains diminish as the dataset size continues to expand. Our findings emphasize two critical axioms in self-supervised learning for computer vision: (1) additional pre-training data can yield diminishing benefits for some datasets and (2) quality is more important than quantity in self-supervised learning, especially when building foundation models.

Recent advancements in behavior cloning have enabled robots to perform complex manipulation tasks. However, accurately assessing training performance remains challenging, particularly for real-world applications, as behavior cloning losses often correlate poorly with actual task success. Consequently, researchers resort to success rate metrics derived from costly and time-consuming real-world evaluations, making the identification of optimal policies and detection of overfitting or underfitting impractical. To address these issues, we propose real-is-sim, a novel behavior cloning framework that incorporates a dynamic digital twin (based on Embodied Gaussians) throughout the entire policy development pipeline: data collection, training, and deployment. By continuously aligning the simulated world with the physical world, demonstrations can be collected in the real world with states extracted from the simulator. The simulator enables flexible state representations by rendering image inputs from any viewpoint or extracting low-level state information from objects embodied within the scene. During training, policies can be directly evaluated within the simulator in an offline and highly parallelizable manner. Finally, during deployment, policies are run within the simulator where the real robot directly tracks the simulated robot's joints, effectively decoupling policy execution from real hardware and mitigating traditional domain-transfer challenges. We validate real-is-sim on the PushT manipulation task, demonstrating strong correlation between success rates obtained in the simulator and real-world evaluations. Videos of our system can be found at https://realissim.rai-inst.com.

Large-scale visuomotor policy learning is a promising approach toward developing generalizable manipulation systems. Yet, policies that can be deployed on diverse embodiments, environments, and observational modalities remain elusive. In this work, we investigate how knowledge from large-scale visual data of the world may be used to address one axis of variation for generalizable manipulation: observational viewpoint. Specifically, we study single-image novel view synthesis models, which learn 3D-aware scene-level priors by rendering images of the same scene from alternate camera viewpoints given a single input image. For practical application to diverse robotic data, these models must operate zero-shot, performing view synthesis on unseen tasks and environments. We empirically analyze view synthesis models within a simple data-augmentation scheme that we call View Synthesis Augmentation (VISTA) to understand their capabilities for learning viewpoint-invariant policies from single-viewpoint demonstration data. Upon evaluating the robustness of policies trained with our method to out-of-distribution camera viewpoints, we find that they outperform baselines in both simulated and real-world manipulation tasks. Videos and additional visualizations are available at https://s-tian.github.io/projects/vista.

A key contributor to recent progress in 3D detection from single images is monocular depth estimation. Existing methods focus on how to leverage depth explicitly, by generating pseudo-pointclouds or providing attention cues for image features. More recent works leverage depth prediction as a pretraining task and fine-tune the depth representation while training it for 3D detection. However, the adaptation is insufficient and is limited in scale by manual labels. In this work, we propose to further align depth representation with the target domain in unsupervised fashions. Our methods leverage commonly available LiDAR or RGB videos during training time to fine-tune the depth representation, which leads to improved 3D detectors. Especially when using RGB videos, we show that our two-stage training by first generating pseudo-depth labels is critical because of the inconsistency in loss distribution between the two tasks. With either type of reference data, our multi-task learning approach improves over the state of the art on both KITTI and NuScenes, while matching the test-time complexity of its single task sub-network.

Building a scalable and real-time recommendation system is vital for many businesses driven by time-sensitive customer feedback, such as short-videos ranking or online ads. Despite the ubiquitous adoption of production-scale deep learning frameworks like TensorFlow or PyTorch, these general-purpose frameworks fall short of business demands in recommendation scenarios for various reasons: on one hand, tweaking systems based on static parameters and dense computations for recommendation with dynamic and sparse features is detrimental to model quality; on the other hand, such frameworks are designed with batch-training stage and serving stage completely separated, preventing the model from interacting with customer feedback in real-time. These issues led us to reexamine traditional approaches and explore radically different design choices. In this paper, we present Monolith, a system tailored for online training. Our design has been driven by observations of our application workloads and production environment that reflects a marked departure from other recommendations systems. Our contributions are manifold: first, we crafted a collisionless embedding table with optimizations such as expirable embeddings and frequency filtering to reduce its memory footprint; second, we provide an production-ready online training architecture with high fault-tolerance; finally, we proved that system reliability could be traded-off for real-time learning. Monolith has successfully landed in the BytePlus Recommend product.

Recent Vision Mamba models not only have much lower complexity for processing higher resolution images and longer videos but also the competitive performance with Vision Transformers (ViTs). However, they are stuck into overfitting and thus only present up to base size (about 80M). It is still unclear how vanilla Vision Mamba (Vim) can be efficiently scaled up to larger sizes, which is essentially for further exploitation. In this paper, we propose a stochastic layer-wise shuffle regularization, which empowers successfully scaling non-hierarchical Vision Mamba to a large size (about 300M) in a supervised setting. Specifically, our base and large-scale ShuffleMamba models can outperform the supervised ViTs of similar size by 0.8\% and 1.0\% classification accuracy on ImageNet1k, respectively, without auxiliary data. When evaluated on the ADE20K semantic segmentation and COCO detection tasks, our ShuffleMamba models also show significant improvements. Without bells and whistles, the stochastic layer-wise shuffle has the following highlights: (1) Plug and play: it does not change model architectures and will be omitted in inference. (2) Simple but effective: it can improve the overfitting in Vim training and only introduce random token permutation operations. (3) Intuitive: the token sequences in deeper layers are more likely to be shuffled as they are expected to be more semantic and less sensitive to patch positions. Code and models will be available at https://github.com/huangzizheng01/ShuffleMamba.

One-shot 3D talking portrait generation aims to reconstruct a 3D avatar from an unseen image, and then animate it with a reference video or audio to generate a talking portrait video. The existing methods fail to simultaneously achieve the goals of accurate 3D avatar reconstruction and stable talking face animation. Besides, while the existing works mainly focus on synthesizing the head part, it is also vital to generate natural torso and background segments to obtain a realistic talking portrait video. To address these limitations, we present Real3D-Potrait, a framework that (1) improves the one-shot 3D reconstruction power with a large image-to-plane model that distills 3D prior knowledge from a 3D face generative model; (2) facilitates accurate motion-conditioned animation with an efficient motion adapter; (3) synthesizes realistic video with natural torso movement and switchable background using a head-torso-background super-resolution model; and (4) supports one-shot audio-driven talking face generation with a generalizable audio-to-motion model. Extensive experiments show that Real3D-Portrait generalizes well to unseen identities and generates more realistic talking portrait videos compared to previous methods. Video samples and source code are available at https://real3dportrait.github.io .