ValueError Traceback (most recent call last)
<ipython-input-80-fbef08da00d7> in <module>
31 #loss = output['loss']
32 #loss = net(img, target['bbox'], target['labels']).to(device)
---> 33 loss = criterion(output, clas).to(device)
34 loss_bb = criterion_bb(outputs, box).to(device)
35 loss.backward()
/opt/conda/lib/python3.7/site-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
725 result = self._slow_forward(*input, **kwargs)
726 else:
--> 727 result = self.forward(*input, **kwargs)
728 for hook in itertools.chain(
729 _global_forward_hooks.values(),
/opt/conda/lib/python3.7/site-packages/torch/nn/modules/loss.py in forward(self, input, target)
528
529 def forward(self, input: Tensor, target: Tensor) -> Tensor:
--> 530 return F.binary_cross_entropy(input, target, weight=self.weight, reduction=self.reduction)
531
532
/opt/conda/lib/python3.7/site-packages/torch/nn/functional.py in binary_cross_entropy(input, target, weight, size_average, reduce, reduction)
2517 if target.size() != input.size():
2518 raise ValueError("Using a target size ({}) that is different to the input size ({}) is deprecated. "
-> 2519 "Please ensure they have the same size.".format(target.size(), input.size()))
2520
2521 if weight is not None:
ValueError: Using a target size (torch.Size([1, 23])) that is different to the input size (torch.Size([1, 19])) is deprecated. Please ensure they have the same size.
for i, (img, boxes, classes) in enumerate(train_loader):
net.to(device)
img = img.to(device)
box = boxes.to(device)
clas = classes.to(device)
optimizer.zero_grad()
output = net(img)
loss = criterion(output, clas).to(device)
loss_bb = criterion_bb(outputs, box).to(device)
loss.backward()
optimizer.step()
I get an error in loss when I pass tensor with lable and bbox into it. model vgg16_bn / I have 19 classes. I assume that the model or loss takes trying to push all the bboxes into the classifier layer, and each image has a different number of objects. how do i feed to bbox model,
label? or maybe this model is not suitable for multi-class + multi-label classification?