search

D-X-Y / landmark-detection

Four landmark detection algorithms, implemented in PyTorch.
https://xuanyidong.com/assets/projects/TPAMI-2020-SRT.html
MIT License
925 stars 180 forks source link

ModuleNotFoundError: No module named 'xvision' #89

Open Koruvika opened 1 year ago

Koruvika commented 1 year ago

When I try open a pretrained model in here, I get a the error "ModuleNotFoundError: No module named 'xvision'", how can i fix it?

D-X-Y commented 1 year ago

Hi Koruvika, the xvision is located at https://github.com/D-X-Y/landmark-detection/tree/master/SRT/lib/xvision it may due to you did not properly import the libraries in this repo

Koruvika commented 1 year ago

Tks for reply @D-X-Y, i tried training with X-SBR-main.py but it seems the model was overfit from the first epochs, training registration loss decreased while validation registration loss increased. The model I use is a pre-trained PFLD. My dataset i use contains 50k image and 50k frames (from video).

Koruvika commented 1 year ago

I saw the loss in temporal_loss_regression.py image past_loss and future_loss seem to calculate only once the distance between two consecutive frames (video_L is int) instead of calculating the distances of all consecutive frames in the sequence, does this affect training results?