Where is training and validation loss in MaskRCNN? Needed for fine-tuning hyperparameters

vision
1

Hi everyone,

I am a complete beginner in deep learning, and I am trying to use this repo (GitHub - mrlooi/rotated_maskrcnn: Rotated Mask R-CNN: From Bounding Boxes to Rotated Bounding Boxes) which extends MaskRCNN.

I don’t really know which is the best way to fine tune the hyperparameters without under or overfitting. I read that a good point of view is plotting training and validation loss, and finding the number of iterations that make them diverge.

However, I can’t find which variables are those values stored in, so I can’t plot them. Does anybody know it or know what could I do to find them?

I am using GoogleColab and TorchVision=1.1.0.

Thank you all!!