Expected 4D input (got 2D input)

vision
1

1
1746×856 35.5 KB
2
2650×698 14.9 KB

File “main1.py”, line 221, in <module>

reid_trainer.train(epoch, train_loader)

File “/app/train.py”, line 44, in train

self._forward()

File “/app/train.py”, line 76, in _forward

score, feat = self.model(self.data)

File “/opt/conda/lib/python3.6/site-packages/torch/nn/modules/module.py”, line 489, in call

result = self.forward(*input, **kwargs)

File “/app/model.py”, line 125, in forward

feature = self.reduction(triplet_feature)

File “/opt/conda/lib/python3.6/site-packages/torch/nn/modules/module.py”, line 489, in call

result = self.forward(*input, **kwargs)

File “/opt/conda/lib/python3.6/site-packages/torch/nn/modules/container.py”, line 92, in forward

input = module(input)

File “/opt/conda/lib/python3.6/site-packages/torch/nn/modules/module.py”, line 489, in call

result = self.forward(*input, **kwargs)

File “/opt/conda/lib/python3.6/site-packages/torch/nn/modules/batchnorm.py”, line 60, in forward

self._check_input_dim(input)

File “/opt/conda/lib/python3.6/site-packages/torch/nn/modules/batchnorm.py”, line 243, in _check_input_dim

.format(input.dim()))

ValueError: expected 4D input (got 2D input)

I created model based on GitHub - daizuozhuo/batch-dropblock-network: Official source code of "Batch DropBlock Network for Person Re-identification and Beyond" (ICCV 2019)

problem here is that even though I just trained with 224x224 size images (which is fixed size of imagenet) the following error log comes up. which part of the model should I fix?
can anybody tell me where the problem is?? thank you

2

Which line are u getting the error?

3

model.py line 125 is feature = self.reduction(triplet_feature)
something wrong with this line??

4

I cannot fully track it but it’s a simple mistake about the amount of features. I see you are applying a max pool over some features l, therefore that tensor should be 4 dimensional, of size BxCx1x1 then you are squeezing it and getting a 2d tensor. So probably you have a missmatching. Just check the outgoing size of those features