fashion-mnist-pytorch

A simple CNN to classify fashion mnist dataset. The following network provides 90% test accuracy. On the MNIST dataset, the test accuracy is 99.2%.

----------------------------------------------------------------
        Layer (type)               Output Shape         Param #
================================================================
            Conv2d-1           [-1, 32, 26, 26]             832
              ReLU-2           [-1, 32, 26, 26]               0
       BatchNorm2d-3           [-1, 32, 26, 26]              64
         MaxPool2d-4           [-1, 32, 13, 13]               0
            Conv2d-5           [-1, 64, 13, 13]          18,496
              ReLU-6           [-1, 64, 13, 13]               0
       BatchNorm2d-7           [-1, 64, 13, 13]             128
         MaxPool2d-8             [-1, 64, 6, 6]               0
           Dropout-9             [-1, 64, 6, 6]               0
           Linear-10                 [-1, 1024]       2,360,320
             ReLU-11                 [-1, 1024]               0
           Linear-12                   [-1, 10]          10,250
================================================================
Total params: 2,390,090
Trainable params: 2,390,090
Non-trainable params: 0
----------------------------------------------------------------
Input size (MB): 0.00
Forward/backward pass size (MB): 0.83
Params size (MB): 9.12
Estimated Total Size (MB): 9.96
----------------------------------------------------------------

Setup

1. Clone this repository git clone https://github.com/gmuraleekrishna/fashion-mnist-pytorch.git
2. Change directory to fashion-mnist-pytorch.
3. Run pip install requirements.py to install the required pachages.

Training

Run training by running python main.py --tensorboard. Best weights will be saved in fashion-mnist.pth.

Testing

Run testing by running python main.py --test --file fashion-mnist.pth