I’m looking through the source code and documentation of fcn_resnet101 and felt a little bit of confused about the pretraining. Is the resnet backbone pretrained on ImageNet while the classifier is pretrained on the subset of Coco?
I’m looking through the source code and documentation of fcn_resnet101 and felt a little bit of confused about the pretraining. Is the resnet backbone pretrained on ImageNet while the classifier is pretrained on the subset of Coco?