Can
Can
miss pose instance for person_1    error pose idx    error mask     this is just one example, to tell the truth, there...
where is the sample code?
for widerface wider_face_train_bbx_gt.txt Number of bounding box x1, y1, w, h, blur, expression, illumination, invalid, occlusion, pose what index does x1 and y1 start, 0 or 1? for celeba list_landmarks_celeba.txt...
I’m confused about how the threshold can be 0.5 for all situations, no matter what M and N is ? if M and N are big enough, f(i, ˆj) can...
It seems the lmk_faces_idx selected to match 68 2D landmark is not so accurate, which can be viewed by Blender as follows:  ...
https://github.com/Zejun-Yang/AniPortrait/issues/143
Thanks for sharing this work, good insight and inspiring. But I'm unable to perceive improvement of the pretrained model. My inference of E_expression: For images, inputs: just concat same image...
https://github.com/Zejun-Yang/AniPortrait/issues/143
我的研究进展
https://github.com/Zejun-Yang/AniPortrait/issues/143
以下是一些效果展示,包括人脸驱动,语音驱动,表情姿态动作迁移等: https://github.com/Zejun-Yang/AniPortrait/assets/20749514/a4bd2a06-6be8-4ecf-bfa3-337e5dcff4bb https://github.com/Zejun-Yang/AniPortrait/assets/20749514/087b2023-67e9-41bf-bf31-204052889716 https://github.com/Zejun-Yang/AniPortrait/assets/20749514/053533eb-4237-4092-b224-cb62f6b03b59