DiffSynth-Studio/Qwen-Image-Edit-F2P

Qwen-Image-Edit Face Generation Image Model

Model Introduction

This model is based on the Qwen-Image-Edit face-controlled image generation model. Given a cropped facial image as input, it generates full portrait images of the same person.

Result Demonstration

Face	Generated Image 1	Generated Image 2	Generated Image 3	Generated Image 4

Inference Code

git clone https://github.com/modelscope/DiffSynth-Studio.git  
cd DiffSynth-Studio
pip install -e .

from diffsynth.pipelines.qwen_image import QwenImagePipeline, ModelConfig
import torch
from modelscope import snapshot_download, dataset_snapshot_download
from PIL import Image

pipe = QwenImagePipeline.from_pretrained(
    torch_dtype=torch.bfloat16,
    device="cuda",
    model_configs=[
        ModelConfig(model_id="Qwen/Qwen-Image-Edit", origin_file_pattern="transformer/diffusion_pytorch_model*.safetensors"),
        ModelConfig(model_id="Qwen/Qwen-Image", origin_file_pattern="text_encoder/model*.safetensors"),
        ModelConfig(model_id="Qwen/Qwen-Image", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
    ],
    tokenizer_config=None,
    processor_config=ModelConfig(model_id="Qwen/Qwen-Image-Edit", origin_file_pattern="processor/"),
)
snapshot_download("DiffSynth-Studio/Qwen-Image-Edit-F2P", local_dir="models/DiffSynth-Studio/Qwen-Image-Edit-F2P", allow_file_pattern="model.safetensors")
pipe.load_lora(pipe.dit, "models/DiffSynth-Studio/Qwen-Image-Edit-F2P/model.safetensors")
dataset_snapshot_download(
    dataset_id="DiffSynth-Studio/example_image_dataset",
    local_dir="./data/example_image_dataset",
    allow_file_pattern="f2p/qwen_woman_face_crop.png"
)
face_image = Image.open("data/example_image_dataset/f2p/qwen_woman_face_crop.png").convert("RGB")

prompt = "Photography. A young woman wearing a yellow dress stands in a flower field, with a background of colorful flowers and green grass."
image = pipe(prompt, edit_image=face_image, seed=42, num_inference_steps=40, height=1152, width=864)
image.save(f"image.jpg")

Face Auto-Cropping

import torch
from PIL import Image
import numpy as np
from insightface.app import FaceAnalysis
import cv2

class FaceDetector(torch.nn.Module):

    def __init__(self):
        super().__init__()
        providers = ["CUDAExecutionProvider", "CPUExecutionProvider"]
        provider_options = [{"device_id": 0}, {}]
        self.app_640 = FaceAnalysis(name='antelopev2', providers=providers, provider_options=provider_options)
        self.app_640.prepare(ctx_id=0, det_size=(640, 640))
        self.app_320 = FaceAnalysis(name='antelopev2', providers=providers, provider_options=provider_options)
        self.app_320.prepare(ctx_id=0, det_size=(320, 320))
        self.app_160 = FaceAnalysis(name='antelopev2', providers=providers, provider_options=provider_options)
        self.app_160.prepare(ctx_id=0, det_size=(160, 160))

    def _detect_face(self, id_image_cv2):
        face_info = self.app_640.get(id_image_cv2)
        if len(face_info) > 0:
            return face_info
        face_info = self.app_320.get(id_image_cv2)
        if len(face_info) > 0:
            return face_info
        face_info = self.app_160.get(id_image_cv2)
        return face_info

    def crop_face(self, id_image):
        face_info = self._detect_face(cv2.cvtColor(np.array(id_image), cv2.COLOR_RGB2BGR))
        if len(face_info) == 0:
            return None
        else:
            bbox = sorted(face_info, key=lambda x: (x['bbox'][2] - x['bbox'][0]) * (x['bbox'][3] - x['bbox'][1]))[-1]['bbox']
            return id_image.crop(list(map(int, bbox)))


face_detector = FaceDetector()
face_image = face_detector.crop_face(Image.open("image_2.jpg"))
face_image.save("face_crop.jpg")