klintan/pytorch-lanenet

Hi!

yerassylbolatov opened this issue · 0 comments

yerassylbolatov commented

Should dataloader deliver the following output to the model?
torch.Size([16, 3, 256, 512]) - image
torch.Size([16, 1, 256, 512]) - binary lane image
torch.Size([16, 5, 256, 512]) - instances of lanes in separate layers

I coded custom dataloader and tusimple json processing script. Now I want to use the proposed model. And I've got the following error.
Screenshot 2023-04-15 at 13 47 24
It seems to me that the dimensions of the dataloader's outputs is not matching with the model, am I right?