fasterrcnn-pytorch-training-pipeline

fasterrcnn-pytorch-training-pipeline copied to clipboard

Hi,

I need more information about how background class is managed. In smoke dataset, I see background class in yaml file, but in fact no background images are presents.

So, you just added one more class during model net definition or did you do something else? How such background class is managed during training phase? Thanks.

So, all the Torchvision detection models expect one extra background class along with the object classes. The class name is not used in the training but instead used to easily manage the indices, that is, the object class index should start from 1.