search

fudan-generative-vision / hallo

Hallo: Hierarchical Audio-Driven Visual Synthesis for Portrait Image Animation
https://fudan-generative-vision.github.io/hallo/
MIT License
7.74k stars 1.01k forks source link

Data Preprocessing error: #face is invalid: 0 #132

Open Dongzhikang opened 1 month ago

Dongzhikang commented 1 month ago

Hi, thank you for the amazing work. When I run the stage one data processing code

python -m scripts.data_preprocess --input_dir dataset_name/videos --step 1

I got many #face is invalid: 0 errors. In the end, I only extracted frames and audio. My face_mask, sep_face_mask, sep_lip_mask and sep_pose_mask folders are all empty.

I tried different videos, but still didn't work. Could you please help me? Thank you!

--------------update---------------- I found there are several empty face landmarks generated by compute_face_landmarks function. Do we need to make sure the face has to be detected in every frame?

xumingw commented 1 month ago

Could you provide a failure case to my email: matthianxu@gmail.com