load_from error

[htaoruan]

Hello,in mmocr0.6，By specifying load_from = 'https://download.openmmlab.com/mmocr/textdet/dbnet/dbnet_r18_fpnc_sbn_1200e_icdar2015_20210329-ba3ab597.pth', the result appears as follows Warning:

2023-02-22 18:03:35,041 - mmcv - WARNING - The model and loaded state dict do not match exactly

unexpected key in source state_dict: fc.weight, fc.bias

[gaotongxiao]

It's not an error. It occurs because the backbone network is pretrained on image classification tasks, where the last fc layer is required to generate the classification output. However, the fc layer is no longer needed when the backbone network is used to extract features in downstream tasks, and therefore these weights can be safely skipped when loading the checkpoint.

[Vania-Guima]

In the last version of MMOCR, when I try to do transfer learning by specifying load_from, freezing or not the backbone and neck, besides the 'unexpected key in source state_dict: fc.weight, fc.bias' warning, also appear the 'missing keys in source state_dict: backbone.stem.0.weight, backbone.stem.1.weight (...)' warning.
Is everything ok?
I think a guideline or tutorial to do transfer learning will be hepfull.

[gaotongxiao]

No, usually you will need to revise the network architecture to take the stem layer away from the backbone.