search

alibaba / animate-anything

Fine-Grained Open Domain Image Animation with Motion Guidance
https://animationai.github.io/AnimateAnything/
MIT License
779 stars 63 forks source link

Question about the training data #22

Closed YujieOuO closed 10 months ago

YujieOuO commented 10 months ago

Thanks for providing animate_anything_svd_v1.0 model.

May i ask how many videos are used to finetune the svd model to acquire animate_anything_svd_v1.0 ? and maybe the different between svd and svd_mask model is the conv_in module? for the input dimension from 8 to 9.

daizuozhuo commented 10 months ago

I use 100k clips for finetuning.

YujieOuO commented 9 months ago

I use 100k clips for finetuning.

Sorry to bother you again, and I'd like to know if the 100k data used to train SVD-mask is from webvid10M?