the pretrained model is an invalid file about landmark-detection HOT 12 CLOSED

Would you mind to provide more information?

from landmark-detection.

snapshot = torch.load(snapshot,map_location='cpu')

I try load the pretrained model downloaded from the google cloud you provided ,but it does not work

from landmark-detection.

Would you mind to provide how do you place the snapshot file and provide a link to this code"snapshot = torch.load(snapshot,map_location='cpu')"?

from landmark-detection.

I run the san_eval.py on CPU, I modified the
snapshot = torch.load(snapshot)
to
snapshot = torch.load(snapshot,map_location='cpu')

from landmark-detection.

Did you exactly run "python san_eval.py --image ./cache_data/cache/test_1.jpg --model ./snapshots/SAN_300W_GTB_itn_cpm_3_50_sigma4_128x128x8/checkpoint_49.pth.tar --face 819.27 432.15 971.70 575.87 --save_path temp_1.png"?

from landmark-detection.

yes , I put the args into the argparse.ArgumentParser

from landmark-detection.

I just add the CPU support for san_eval.py, and tried, but I did not find your issues.
Please make sure "checkpoint_49.pth.tar" is at snapshots/SAN_300W_GTB_itn_cpm_3_50_sigma4_128x128x8

from landmark-detection.

I just replace the san_eval.py you modified ,it still does not work ,the problem is all the same

from landmark-detection.

Can you just share a screenshot of your error?

from landmark-detection.

from landmark-detection.

It might be caused by the PyThon version. I use Python 3.6 to save the model, but you are using Python 3.5. I also have this problem sometimes when load a model saved in different Python version. Would you mind to try Python 3.6?

from landmark-detection.

tks, I have run it on Python3.6 and it really works.

from landmark-detection.