Running transducive-vos-pytorch in the docker container returns torch not found and mask issue

[kausiksivakumar]

Running python transductive-vos.pytorch/run_video.py --img_dir [THE PATH TO COLOR FILES] --init_mask_file [THE INITIAL MASK FILE YOU PREPARED ABOVE] --mask_save_dir [WHERE TO SAVE] to generate masks returns that torch not found error in the container. If I try to install torch it screws up the numpy version due to which I am unable to proceed.

I also wanted to ask what these mask represent. I thought that the BundleTrack code only requires mask to be at time t=0 (initially) but when I look at the dataset I found that masks were available in plenty for each image, could you help me understand this. Thanks

[kausiksivakumar]

If anyone is stuck with the same issue, install pytorch inside the container and do pip install -U numpy . This solves the issue