Bounding Box Regression

[SamihaSara]

Does this version of FasterRCNN do bounding box regression? Could you please point where I can find the related codes? If it is not implemented already, could you please consider adding this feature? Thanks.

[oke-aditya]

Hello @SamihaSara ! Thank you for brining this question!

The FasterRCNN used here is from torchvision.
You can find FasterRCNN code here !

Bounding Box regression: -

It is step where.
A Multi Latered Perceptron is trying to minimize square error loss between predicted boxes and then original boxes.

Yes this is implemented in torchvision, you can read the docs here.

FasterRCNN does give you multiple loss outputs, regression loss, classification loss.
These are returned as metrics from train_step! (averaged to all images in epoch)

Have a look here

Thanks a lot for asking. Hope you like the library.