search

TiagoCortinhal / SalsaNext

Uncertainty-aware Semantic Segmentation of LiDAR Point Clouds for Autonomous Driving
MIT License
413 stars 102 forks source link

Error trying to train from pre-trained weights ''Missing key(s) in state_dict:'' #65

Closed DavideCoppola97 closed 3 years ago

DavideCoppola97 commented 3 years ago

Hi, when I try to start a training using the weights you published I get this error: Trying to train from scratch on the other hand, everything seems to be fine The platform is Windows 10 with Anaconda I have made sure that the directory is right and in fact the SalsaNext file is correctly opened, however it continues to generate this error, as if expecting something different in the content, it seems that the modules are all there, however the script expects to find only the module names while in the file there is the suffix '.module' Cattura Cattura2

DavideCoppola97 commented 3 years ago

I solved it by removing the prefix `` module. '' from the dictionary with a simple for.

FUTUREEEEEE commented 2 years ago

simply Comment out this line https://github.com/Halmstad-University/SalsaNext/blob/a02fad97d646d4c132266ab79fbaea3ecfc237ed/train/tasks/semantic/modules/user.py#L66