wrong predicted symmetry and graph

[fbas-est]

Greetings,

I runned recently the test part of the coretrainer.py with your pre-trained weights and I got wrong outputs for symmetry and graph compared to the outputs provided by you, as you can see in the attached files.

Any idea why this is happening?

your_output.zip
network_output.zip

[chensong1995]

Hello fbas-est,

Thanks for your interest in our work! From what I can see, the keypoint predictions are pretty accurate, but the graph and symmetry are completely off. This is pretty strange because the three itermediate representations are predicted by the same network. My suggestion is to use a debugger and trace the code line by line from the network prediction to the final output. If we can locate the exact line of code that causes the problem, we can start from there. I hope this helps!

Best,
Chen