TMElyralab / MusePose

MusePose: a Pose-Driven Image-to-Video Framework for Virtual Human Generation
Other
1.87k stars 125 forks source link

请问相比于Moore-AnimateAnyone,在哪些方面做了改进? #9

Open BugsMaker0513 opened 1 month ago

TZYSJTU commented 1 month ago

Thanks for your interest in our project! As we claimed in the readme, we fixed several important bugs and made some improvements based on the code of Moore. Please compare the differences between the codes for details. Besides, we could never know the data quality of Moore's dataset.

shensheng272 commented 1 month ago

1、RandomResizedCrop -> Resize, sample rate 2、Motion_module requiresgrad() 3、from_pretrained_2d part load weight 4、uniform closed_loop set to false

Which one is more important? I would assume it's the dataloader.

ZhangLei999 commented 1 month ago

@shensheng272 from_pretrained_2d part load weight ? what's mean? Can you tell me the name of py? Thanks

shensheng272 commented 1 month ago

@shensheng272 from_pretrained_2d part load weight ? what's mean? Can you tell me the name of py? Thanks

src/models/unet_3d.py

BugsMaker0513 commented 4 weeks ago

1、RandomResizedCrop -> Resize, sample rate 2、Motion_module requiresgrad() 3、from_pretrained_2d part load weight 4、uniform closed_loop set to false

Which one is more important? I would assume it's the dataloader.

2、Motion_module requiresgrad() 4、uniform closed_loop set to false Can you tell me the name of py? Thanks