README

This is the code of Ferry_Li for the 4th DataCV Challenge in conjunction with ICCV workshop 2025: Hybrid Generative Fusion for Efficient and Privacy-Preserving Face Recognition Dataset Generation

FRAMEWORK

Dataset

The 10K-scale training set is here.

Steps

1. Step 1: Use feature_extractor.py provided in https://github.com/HaiyuWu/Vec2Face to extract image embeddings from the current HSFace dataset.

2. Step 2: Run reduce_hsface.py to record only consistent identity image paths in the HSFace dataset.

3. Step 3: Run gpt_clean_parallel.py to record inconsistent identity image paths in the HSFace dataset.

4. Step 4: Run convert.py to convert the json format to txt format.

5. Step 5: Run hsface_makeup.py to augment images to 50 per identity, and save the augmented images.

6. Step 6: Run generate_id.py to generate a new image for one identity based on various prompts.

7. Step 7: Use image_generation_with_reference.py to expand one image to 50 per identity.

8. Step 8: Run merge_dataset.py to merge the cleaned HSFace dataset and diffusion-Vec2Face-expanded dataset.