alexhex7 / capsnet_pytorch

Dynamic Routing Between Capsules. Refference [naturomics&timomernick]

CapsNet_pytorch

• Refer to naturomics/CapsNet-Tensorflow(Tensorflow) and timomernick/pytorch-capsule(Pytorch).
• Implementation of Dynamic Routing Between Capsules.

Statement

• Do the Experiments on the MNIST dataset.
• The accuracy and loss records can be find in nohup.out.
• The reconstruction img can be find in result_img.

Append

• CapsulesNet 的解析及整理
• Capsule Networks Explained
• Capsule Networks Are Shaking up AI — Here’s How to Use Them

Reconstruction Image

capsnet_pytorch's Issues

Error on RGB images or channel 3

I want to try to implement your code on RGB images where I got the following errors. Can you help me to sort out the problem please?

transform = transforms.Compose([
transforms.Resize((28,28)),
transforms.Grayscale(num_output_channels=3),
transforms.ToTensor(),
transforms.Normalize((0.1307,), (0.3081,))])

File "main.py", line 117, in
predict, reconstruct_img = net(img_batch, label_batch, train=True)
File "/home/user/pytorch_python3/lib/python3.5/site-packages/torch/nn/modules/module.py", line 224, in call
result = self.forward(*input, **kwargs)
File "/media/user/DATA/New_CODE/Working/CapsNet_pytorch/lib/network.py", line 45, in forward
output = self.conv1(x)
File "/home/user/pytorch_python3/lib/python3.5/site-packages/torch/nn/modules/module.py", line 224, in call
result = self.forward(*input, **kwargs)
File "/home/user/pytorch_python3/lib/python3.5/site-packages/torch/nn/modules/conv.py", line 254, in forward
self.padding, self.dilation, self.groups)
File "/home/user/pytorch_python3/lib/python3.5/site-packages/torch/nn/functional.py", line 52, in conv2d
return f(input, weight, bias)
RuntimeError: Need input.size[1] == 1 but got 3 instead.

Error on USPS dataset.

I would like to thank you so much for this code. You did a great job.

Actually I wanted to implement this code on USPS dataset. I am using the data loading code.

https://github.com/corenel/torchzoo/blob/master/torchzoo/datasets/usps.py

USPS

train_dataset = USPS('/media/DL/DATA/New_CODE/Working/CapsNet_pytorch/USPS', train=True, transform=None, download=False)
test_dataset = USPS('/media/DL/DATA/New_CODE/Working/CapsNet_pytorch/USPS', train=False, transform=None, download=False)
But got the following error.
RuntimeError: Need input.size[1] == 1 but got 28 instead.

Can you help me to sort out the problem please? Thanks in advanced.