wrong image size

[lelechen63]

Thanks for your excellent work! I am trying to follow your instruction and test your sample. Here is the erro mesage I got. I am using your testing sample. Any suggeston? Thanks in advance.

[lelechen63]

@XYWang-coder Hi there, anyone can take a look at this problem? Thanks!

[XYWang-coder]

Hi, I think you can check the shape of the input_rgb in the PENet.py. The input image shape of the ProxyEstimationNet is (3, 200, 150).

[lelechen63]

Hi, I think you can check the shape of the input_rgb in the PENet.py. The input image shape of the ProxyEstimationNet is (3, 200, 150).

Hi there, thansk fo your quick reply. I am using your sample imges as input. The images in /sample folder are size of 3x1200x1600. The input img to the model_NormalEst in file main.py line41 is size of [3,3,800,600]. The images to the ProxyEstimationNet is size of ([1, 3, 200, 150]). I still have the same error message.

[XYWang-coder]

It seems to be something wrong with the avgpool of the resnet18. I checked the input size and output size of every layer and the information is shown in the following.

You can also check this information in the PENet.py by adding code like

[lelechen63]

It seems to be something wrong with the avgpool of the resnet18. I checked the input size and output size of every layer and the information is shown in the following.

You can also check this information in the PENet.py by adding code like

@XYWang-coder can you add my wechat to solve this problem？ Thanks, my wechat is： cll1_2_3_4_5cll

[lelechen63]

It seems to be something wrong with the avgpool of the resnet18. I checked the input size and output size of every layer and the information is shown in the following.

You can also check this information in the PENet.py by adding code like

@XYWang-coder can you add my wechat to solve this problem？ Thanks, my wechat is： cll1_2_3_4_5cll

You are rright, it is pooling's problem.

[XYWang-coder]

I think you can find some information about resnet18 in Pytorch to solve this problem.

[zarkua]

hi @XYWang-coder @lelechen63 i have the same mistake! Is there a way to fix it?

[zarkua]

@Juyong @XYWang-coder anyone can take a look at this problem? Thanks!

[XYWang-coder]

Well this problem may be caused by 'avgpool' operation in pytorch. I think you can start from this aspect to find a solution. I found a similar question and some discussions here pytorch/vision#696, hope it helps.