fasterrcnn-pytorch-training-pipeline

fasterrcnn-pytorch-training-pipeline copied to clipboard

Hi sovit, its been a long time. I finally released from my finals. Now that I would like to continue to study this model, so that it could output the validation loss graph (because my thesis need it to check if the model overfit or underfit when applying it to an unseen dataset).

So, as last time, I had found a website: How can I determine validation loss for faster RCNN (PyTorch)?

And you have said it might required to break the model to several different module. So can you provide some guidance like how to break the model or how to implement the method used in the website?

@MheadHero There are a few ways. But there are other important things that you may want to consider before moving into the validation loss path. One thing is that in object detection, you can also check out overfitting using the mAP that is shown after each epoch for the validation dataset.

What I am asking is that before working on the validation loss part, you may want to discuss this with your teammates/guide and then finalize it. The reason I am saying this is because validation mAP can also be used to check whether a model is overfitting or not and following the standard method you will be able to make other major/architectural changes to the model down the line which may be beneficial.