Update README.md

README.md

@@ -130,36 +130,40 @@ configs:
     path: data/test-*
 ---
 
-# Bird3m Dataset
 
-## Dataset Description
-
-This dataset contains multi-modal data for bird tracking and behavior analysis, primarily focused on Zebrafinches (based on category names in the source data). Each data entry corresponds to a specific bird instance within a video frame.
+# Bird3M Dataset
 
-The data includes:
+## Dataset Description
 
-*   3D keypoints derived from multi-view reconstruction.
-*   2D keypoints and bounding boxes for top, side, and back camera views.
-*   Information about the video frame and associated processed audio/radio data files.
-*   Metadata about the bird and the experimental setup.
-*   Linked vocalization events associated with the specific bird in the frame.
+**Bird3M** is the first synchronized, multi-modal, multi-individual dataset designed for comprehensive behavioral analysis of freely interacting birds, specifically zebra finches, in naturalistic settings. It addresses the critical need for benchmark datasets that integrate precisely synchronized multi-modal recordings to support tasks such as 3D pose estimation, multi-animal tracking, sound source localization, and vocalization attribution. The dataset facilitates research in machine learning, neuroscience, and ethology by enabling the development of robust, unified models for long-term tracking and interpretation of complex social behaviors.
 
-This dataset is designed to facilitate research in areas such as multi-view 3D reconstruction, multi-modal data fusion, and fine-grained animal behavior analysis.
+### Key Features
+- **Duration**: 22.2 hours of synchronized multi-modal recordings, including a fully annotated subset with 4,420 video frames and 2.5 hours of contextual audio and sensor data.
+- **Modalities**:
+  - **Multi-view video**: Three orthogonal color cameras (top, side, back) at 47 fps, supplemented by a monochrome nest camera.
+  - **Multi-channel audio**: Wall-mounted microphones (16 kHz) and body-mounted accelerometers (24,414 Hz, down-sampled to 16 kHz).
+  - **Radio signals**: FM radio phases and magnitudes from four orthogonal antennas.
+- **Annotations**:
+  - **Visual**: 57,396 3D keypoints (5 per bird: beak tip, head center, backpack center, tail base, tail end) across 4,420 frames, with 2D keypoints, visibility labels, and bounding boxes.
+  - **Audio**: 4,902 vocalization segments with onset/offset times and vocalizer identities, linked across microphone and accelerometer channels.
+- **Experimental Setup**: Data from 15 experiments (2–8 birds each) conducted in the **Birdpark** system, with sessions lasting 4–120 days.
+- **Applications**: Supports 3D localization, pose estimation, multi-animal tracking, sound source localization/separation, and cross-modal behavioral analyses (e.g., vocalization directness).
 
-[More details about the dataset origin, collection process, and purpose can be added here.]
+### Purpose
+Bird3M bridges the gap in publicly available datasets for multi-modal animal behavior analysis by providing:
+1. A benchmark for unified machine learning models tackling multiple behavioral tasks.
+2. A platform for exploring efficient multi-modal information fusion.
+3. A resource for ethological studies linking movement, vocalization, and social context to uncover neural and evolutionary mechanisms.
 
 ## Dataset Structure
 
-The dataset is structured into splits based on the `split` field in the original data. The standard splits are `train`, `val` (mapped from `val`), and `test`.
-
-Each split is a standard Hugging Face `Dataset` object. Each row in the dataset corresponds to a single detected bird instance in a single frame, with associated multi-modal data.
+The dataset is organized into three splits: `train`, `val`, and `test`, each as a Hugging Face `Dataset` object. Each row corresponds to a single bird instance in a video frame, with associated multi-modal data.
 
+### Accessing Splits
 ```python
-# Example of accessing splits
 from datasets import load_dataset
 
-dataset = load_dataset("anonymous-submission000/bird3m") # Replace with your actual repo_id
-
+dataset = load_dataset("anonymous-submission000/bird3m")
 train_dataset = dataset["train"]
 val_dataset = dataset["val"]
 test_dataset = dataset["test"]
@@ -167,53 +171,52 @@ test_dataset = dataset["test"]
 
 ## Dataset Fields
 
-Each example in the dataset has the following fields:
+Each example includes the following fields:
 
-*   `bird_id` (`string`): A unique identifier for the specific bird instance within the context of its frame (e.g., "bird_1", "bird_2").
-*   `back_bbox_2d` (`Sequence[float64]`): 2D bounding box coordinates for the bird instance in the **back** camera view. Format is likely `[x, y, width, height]` or `[x_min, y_min, x_max, y_max]`.
-*   `back_keypoints_2d` (`Sequence[float64]`): 2D keypoint coordinates and visibility flags for the bird instance in the **back** camera view. Format is likely `[x1, y1, v1, x2, y2, v2, ...]`, where `v` is the visibility status (e.g., 0: not labeled, 1: labeled but not visible, 2: visible and labeled).
-*   `back_view_boundary` (`Sequence[int64]`): Boundary coordinates defining the relevant area for the **back** view within the full image dimensions. Format is likely `[x, y, width, height]`.
-*   `bird_name` (`string`): The biological or specific identifier assigned to the bird (e.g., "b13k20_f", "b13o15_m", "dead", "2U7a_j").
-*   `video_name` (`string`): Identifier for the original video file the frame belongs to (e.g., "BP_2020-10-13_19-44-38_564726_0240000").
-*   `frame_name` (`string`): Filename of the individual frame (e.g., "img00961.png").
-*   `frame_path` (`Image`): Path to the image file (`.png`) for this specific frame. When accessing this field using the `datasets` library, it will automatically load the image.
-*   `keypoints_3d` (`Sequence[Sequence[float64]]`): 3D keypoint coordinates for the bird instance. Each inner sequence is a 3D point, likely `[x, y, z]`.
-*   `radio_path` (`Value(dtype='binary')`): Path to an associated radio data file (`.npz`). Stored as generic binary data. You would need external libraries (like `numpy`) to load/interpret the content of the `.npz` file after reading the path.
-*   `reprojection_error` (`Sequence[float64]`): Reprojection error values, likely corresponding to each 3D keypoint.
-*   `side_bbox_2d` (`Sequence[float64]`): 2D bounding box for the **side** camera view. (Format similar to `back_bbox_2d`).
-*   `side_keypoints_2d` (`Sequence[float64]`): 2D keypoint coordinates and visibility for the **side** camera view. (Format similar to `back_keypoints_2d`).
-*   `side_view_boundary` (`Sequence[int64]`): View boundary coordinates for the **side** view. (Format similar to `back_view_boundary`).
-*   `backpack_color` (`string`): Color of the backpack tag on the bird (e.g., "purple", "yellow", "red").
-*   `experiment_id` (`string`): Simplified experiment identifier (e.g., "copExpBP03", "juvExpBP05").
-*   `split` (`string`): Dataset split for this example ("train", "val", "test").
-*   `top_bbox_2d` (`Sequence[float64]`): 2D bounding box for the **top** camera view. (Format similar to `back_bbox_2d`).
-*   `top_keypoints_2d` (`Sequence[float64]`): 2D keypoint coordinates and visibility for the **top** camera view. (Format similar to `back_keypoints_2d`).
-*   `top_view_boundary` (`Sequence[int64]`): View boundary coordinates for the **top** view. (Format similar to `back_view_boundary`).
-*   `video_path` (`Video`): Path to the video clip file (`.mp4`) containing this frame. When accessing this field, it will automatically load the video object.
-*   `acc_ch_map` (`{'0': string, ...}`): Dictionary mapping accelerometer channel indices (as strings) to bird identifiers or descriptions.
-*   `acc_sr` (`float64`): Accelerometer sampling rate in Hz.
-*   `has_overlap` (`bool`): Boolean indicating whether the accelerometer event overlaps with the vocalization event (based on previous modification logic).
-*   `mic_ch_map` (`{'0': string, ...}`): Dictionary mapping microphone channel indices (as strings) to microphone names or descriptions.
-*   `mic_sr` (`float64`): Microphone sampling rate in Hz.
-*   `acc_path` (`Audio`): Path to the processed accelerometer audio file (`.wav`). When accessing this field, it will automatically load the audio signal.
-*   `mic_path` (`Audio`): Path to the processed microphone audio file (`.wav`). When accessing this field, it will automatically load the audio signal.
-*   `vocalization` (`Sequence[Dict]`): A list (Sequence) containing dictionaries (Dict) representing vocalization events associated with this bird in this frame. Each dictionary within this list has the following fields:
-    *   `overlap_type` (`string`): Describes the type or confidence of overlap/attribution (mapped from `attribution_confidence_step2`).
-    *   `has_bird` (`bool`): Boolean indicating if this vocalization event was attributed to a bird (mapped from `is_attributed_to_bird`).
-    *   `2ddistance` (`bool`): Boolean indicating if the 2D keypoint match distance for this vocalization attribution was within a certain threshold (likely < 20px, mapped from `keypoint_match_is_close_lt_20px`).
-    *   `small_2ddistance` (`float64`): The minimum 2D keypoint match distance associated with this vocalization event attribution (mapped from `keypoint_match_min_distance_px`).
-    *   `voc_metadata` (`Sequence[float64]`): Likely the onset and offset times of the vocalization event within the associated audio clip (e.g., `[onset_sec, offset_sec]`). Mapped from `vocalization_onset_offset_sec_in_clip`.
+- **`bird_id`** (`string`): Unique identifier for the bird instance (e.g., "bird_1").
+- **`back_bbox_2d`** (`Sequence[float64]`): 2D bounding box for the back view, format `[x_min, y_min, x_max, y_max]`.
+- **`back_keypoints_2d`** (`Sequence[float64]`): 2D keypoints for the back view, format `[x1, y1, v1, x2, y2, v2, ...]`, where `v` is visibility (0: not labeled, 1: labeled but invisible, 2: visible).
+- **`back_view_boundary`** (`Sequence[int64]`): Back view boundary, format `[x, y, width, height]`.
+- **`bird_name`** (`string`): Biological identifier (e.g., "b13k20_f").
+- **`video_name`** (`string`): Video file identifier (e.g., "BP_2020-10-13_19-44-38_564726_0240000").
+- **`frame_name`** (`string`): Frame filename (e.g., "img00961.png").
+- **`frame_path`** (`Image`): Path to the frame image (`.png`), loaded as a PIL Image.
+- **`keypoints_3d`** (`Sequence[Sequence[float64]]`): 3D keypoints, format `[[x1, y1, z1], [x2, y2, z2], ...]`.
+- **`radio_path`** (`binary`): Path to radio data (`.npz`), stored as binary.
+- **`reprojection_error`** (`Sequence[float64]`): Reprojection errors for 3D keypoints.
+- **`side_bbox_2d`** (`Sequence[float64]`): 2D bounding box for the side view.
+- **`side_keypoints_2d`** (`Sequence[float64]`): 2D keypoints for the side view.
+- **`side_view_boundary`** (`Sequence[int64]`): Side view boundary.
+- **`backpack_color`** (`string`): Backpack tag color (e.g., "purple").
+- **`experiment_id`** (`string`): Experiment identifier (e.g., "CopExpBP03").
+- **`split`** (`string`): Dataset split ("train", "val", "test").
+- **`top_bbox_2d`** (`Sequence[float64]`): 2D bounding box for the top view.
+- **`top_keypoints_2d`** (`Sequence[float64]`): 2D keypoints for the top view.
+- **`top_view_boundary`** (`Sequence[int64]`): Top view boundary.
+- **`video_path`** (`Video`): Path to the video clip (`.mp4`), loaded as a Video object.
+- **`acc_ch_map`** (`struct`): Maps accelerometer channels to bird identifiers.
+- **`acc_sr`** (`float64`): Accelerometer sampling rate (Hz).
+- **`has_overlap`** (`bool`): Indicates if accelerometer events overlap with vocalizations.
+- **`mic_ch_map`** (`struct`): Maps microphone channels to descriptions.
+- **`mic_sr`** (`float64`): Microphone sampling rate (Hz).
+- **`acc_path`** (`Audio`): Path to accelerometer audio (`.wav`), loaded as an Audio signal.
+- **`mic_path`** (`Audio`): Path to microphone audio (`.wav`), loaded as an Audio signal.
+- **`vocalization`** (`list[struct]`): Vocalization events, each with:
+  - `overlap_type` (`string`): Overlap/attribution confidence.
+  - `has_bird` (`bool`): Indicates if attributed to a bird.
+  - `2ddistance` (`bool`): Indicates if 2D keypoint distance is <20px.
+  - `small_2ddistance` (`float64`): Minimum 2D keypoint distance (px).
+  - `voc_metadata` (`Sequence[float64]`): Onset/offset times `[onset_sec, offset_sec]`.
 
 ## How to Use
 
+### Loading and Accessing Data
 ```python
 from datasets import load_dataset
-import numpy as np # Needed to load .npz radio data
+import numpy as np
 
-# Load the dataset from the Hub
-dataset = load_dataset("anonymous-submission000/bird3m") # Replace with your actual repo_id
-
-# Access a split
+# Load dataset
+dataset = load_dataset("anonymous-submission000/bird3m")
 train_data = dataset["train"]
 
 # Access an example
@@ -223,32 +226,30 @@ example = train_data[0]
 bird_id = example["bird_id"]
 keypoints_3d = example["keypoints_3d"]
 top_bbox = example["top_bbox_2d"]
-vocalizations = example["vocalization"] # This is a list of dicts
-
-# Access multimedia files (they are lazy-loaded)
-image = example["frame_path"] # This loads the PIL Image
-video = example["video_path"] # This loads a Video object
-mic_audio = example["mic_path"] # This loads the Audio signal (dict with 'array' and 'sampling_rate')
-acc_audio = example["acc_path"] # This loads the Audio signal
+vocalizations = example["vocalization"]
 
-# To access the audio arrays and sampling rates:
-mic_array = mic_audio['array']
-mic_sr = mic_audio['sampling_rate']
-acc_array = acc_audio['array']
-acc_sr_actual = acc_audio['sampling_rate'] # Note: This comes from the file metadata, compare with example['acc_sr']
+# Load multimedia
+image = example["frame_path"]  # PIL Image
+video = example["video_path"]  # Video object
+mic_audio = example["mic_path"]  # Audio signal
+acc_audio = example["acc_path"]  # Audio signal
 
-# To access the binary radio data (if you know how to parse it, e.g., with numpy)
-radio_bytes = example["radio_path"] # This loads the binary content
-# Example (assuming it's a numpy .npz file):
-# try:
-#     from io import BytesIO
-#     radio_data = np.load(BytesIO(radio_bytes))
-#     # Access data inside the npz file, e.g., radio_data['some_key']
-#     print("Radio data keys:", list(radio_data.keys()))
-# except Exception as e:
-#     print(f"Could not load radio data: {e}")
+# Access audio arrays
+mic_array = mic_audio["array"]
+mic_sr = mic_audio["sampling_rate"]
+acc_array = acc_audio["array"]
+acc_sr = acc_audio["sampling_rate"]
 
+# Load radio data
+radio_bytes = example["radio_path"]
+try:
+    from io import BytesIO
+    radio_data = np.load(BytesIO(radio_bytes))
+    print("Radio data keys:", list(radio_data.keys()))
+except Exception as e:
+    print(f"Could not load radio data: {e}")
 
+# Print example info
 print(f"Bird ID: {bird_id}")
 print(f"Number of 3D keypoints: {len(keypoints_3d)}")
 print(f"Top Bounding Box: {top_bbox}")
@@ -256,16 +257,28 @@ print(f"Number of vocalization events: {len(vocalizations)}")
 
 if vocalizations:
     first_vocal = vocalizations[0]
-    print(f"First vocal event metadata: {first_vocal.get('voc_metadata')}")
-    print(f"First vocal event overlap type: {first_vocal.get('overlap_type')}")
+    print(f"First vocal event metadata: {first_vocal['voc_metadata']}")
+    print(f"First vocal event overlap type: {first_vocal['overlap_type']}")
+```
 
-# Example of how to get a slice of the audio clip based on voc_metadata
-# if vocalizations and mic_sr is not None:
-#     onset, offset = vocalizations[0]['voc_metadata']
-#     onset_sample = int(onset * mic_sr)
-#     offset_sample = int(offset * mic_sr)
-#     vocal_audio_clip = mic_array[onset_sample:offset_sample]
-#     print(f"Duration of first vocal clip: {offset - onset:.3f} seconds")
-#     print(f"Shape of first vocal audio clip: {vocal_audio_clip.shape}")
+### Example: Extracting Vocalization Audio Clip
+```python
+if vocalizations and mic_sr:
+    onset, offset = vocalizations[0]["voc_metadata"]
+    onset_sample = int(onset * mic_sr)
+    offset_sample = int(offset * mic_sr)
+    vocal_audio_clip = mic_array[onset_sample:offset_sample]
+    print(f"Duration of first vocal clip: {offset - onset:.3f} seconds")
+    print(f"Shape of first vocal audio clip: {vocal_audio_clip.shape}")
+```
+**Code Availability**: Baseline code is available at [https://github.com/anonymoussubmission0000/bird3m](https://github.com/anonymoussubmission0000/bird3m).
 
+## Citation
+```bibtex
+@article{2025bird3m,
+  title={Bird3M: A Multi-Modal Dataset for Social Behavior Analysis Tool Building},
+  author={tbd},
+  journal={arXiv preprint arXiv:XXXX.XXXXX},
+  year={2025}
+}
 ```