This is a MNIST number recognition problem solved by a convolution neural network model.
Implement by tensorflow 0.11
python2.7
tensorflow 0.11
progressbar
numpy
Download MNIST training data and its labels from here
train-images-idx3-ubyte.gz: training set images (9912422 bytes)
train-labels-idx1-ubyte.gz: training set labels (28881 bytes)
Place "train-images-idx3-ubyte" and "train-labels-idx1-ubyte" in the same directory as this project
$ python cnn.py
##Performance
Train/Dev: 60000/2000
| Data | Accuarrcy |
|---|---|
| train | 0.998 |
| dev | 0.995 |
This code will automatically dump the prediction output of the test-image
TODO: image distortion