search

amjltc295 / Free-Form-Video-Inpainting

Official Pytorch implementation of "Learnable Gated Temporal Shift Module for Deep Video Inpainting. Chang et al. BMVC 2019." and the FVI dataset in "Free-form Video Inpainting with 3D Gated Convolution and Temporal PatchGAN, Chang et al. ICCV 2019"
https://arxiv.org/abs/1907.01131
335 stars 53 forks source link

GPU memory #16

Closed AmingWu closed 4 years ago

AmingWu commented 4 years ago

Could I use a 1080ti GPU to train your model? Thank you

amjltc295 commented 4 years ago

Yes, you could adjust the batch size and video length. However, the performance would not be as good.

AmingWu commented 4 years ago

If I want to reproduce your result, how much GPU memory should I need? and how long your model needs to train?

amjltc295 commented 4 years ago

We use 1-2 V100s with 16G RAM. It takes 1-3 days for LGTSM and 3-10 days for 3DGated including fine-tuning with GAN.