eccv2020

copy /u/lchen63/Project/face_tracking_detection/eccv2020/basics under this folder. We need some dependences in basics forder.
The video frame cropper for talking head is in ./data/face_tracker.py line 72 _crop_video(video, pid = 0) function.

For data preprocess:

data/single_video_preprocess.py

use landamrk_extractor to extract landmarks.
use computeRT to compute the RT betwen the canonical landmark and target landmark

use get_front_video to find the front face to represent the video.

modify get_3d and copy it to ~/github/PRNet and run under py36 environemnt.

go to pytorch1.1python3 conda environment,
use face_tool find_camera to generate rendered 3D video.