search

lsa1997 / POP

Code for "Learning Orthogonal Prototypes for Generalized Few-shot Semantic Segmentation" [CVPR2023]
MIT License
20 stars 3 forks source link

resnet50 #7

Open jwx15 opened 4 months ago

jwx15 commented 4 months ago

Hello, the ResNet50 network you are using returns a list, while using ResNet50V2 returns a tensor. However, if not properly processed, it defaults to using the network output from ResNet50V2, and an error will occur if ResNet50 is used. Also, your backbone was not frozen during training, right? Also, I can't open the link you provided to get the pretrained ResNet50 model. Do you have any other solutions?

lsa1997 commented 4 months ago

The pre-trained models are initially provided by semseg, so you can download them here.