CCAM
CCAM copied to clipboard
Is it true that the encoder will be trained from scratch in WSSS?
Is is true that the backbone will also be retrained from scratch in the WSSS task?
Hi, we initialized the encoder with supervised/unsupervised pretrained weights then retrained it using PASCAL VOC for WSSS.
Thank you so much for your response. Have your team conducted the experiment without retraining the backbone but only the decoding part?
Hi, we didn't try that.