Zhong0x29a/MNIST-Classifier

MNIST-Classifier

Prerequisites

System requirements

• Python 3
• CPU or NVIDIA GPU + CUDA

Dependencies

• torch >= 1.0.0

Prepare Dataset

• Make Folder

  mkdir ./dataset/

• Download

  wget http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz -O ./dataset/train-images-idx3-ubyte.gz
  wget http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz -O ./dataset/train-labels-idx1-ubyte.gz
  wget http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz -O ./dataset/t10k-images-idx3-ubyte.gz
  wget http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz -O ./dataset/t10k-labels-idx1-ubyte.gz

• Unpack

  gzip -d ./dataset/train-images-idx3-ubyte.gz
  gzip -d ./dataset/train-labels-idx1-ubyte.gz
  gzip -d ./dataset/t10k-images-idx3-ubyte.gz
  gzip -d ./dataset/t10k-labels-idx1-ubyte.gz

Usage

• 1.Train Use:

  python main.py train

  to train model, and your model will be saved to ./model.pth

• 2.Test Use:

  python main.py test

  to test your model via the test dataset.

• 3.Recognize your own images Use:

  python main.py recognize

  to recognize your .bmp images in the path (./unlabeled/img_name.bmp) (the images must be 28*28 in size, and with 256 channels in black&white), the results will be been both in console and the path (./labeled/result_img_name.bmp) (the 'result' refers to the recognized number).