search

taohan10200 / IIM

PyTorch implementations of the paper: "Learning Independent Instance Maps for Crowd Localization"
MIT License
163 stars 39 forks source link

net.load_state_dict(new_state_dict,False) #17

Open henbucuoshanghai opened 3 years ago

henbucuoshanghai commented 3 years ago

size mismatch for Binar.Threshold_Module.0.weight: copying a param with shape torch.Size([256, 768, 3, 3]) from checkpoint, the shape in current model is torch.Size([256, 720, 3, 3]).

why？

taohan10200 commented 3 years ago

I guess that you mismatch the model with the pre-trained parameters. We provide two models in IIM, VGG16-FPN (output 720 channels after the FPN layer) and HRNET (output 768 channels). Please check which model you are calling and which parameters you are using.