Awesome

handwritten-digits-recognition

Handwritten digits recognition using NN and Ruby 🧠💎

About the project

This project is using neural network written in ruby for handwritten digits recognition. It is complete solution with frontend and backend parts. Model was trained and tested on MNIST handwritten digits set converted into libsvm format dataset. Set contains 60k train and 10k test examples.

digits 0-9

Instalation

Application is using Rumale::Torch, which requires to install LibTorch:

Running app with docker

Build image:

docker build -t hdr_app .

Run container:

docker run -p 4567:4567 hdr_app:latest

Running app locally

For MacOS:

brew install automake libtorch

Then execute:

bundle install

For Linux please follow Dockerfile when installing libtorch.

Model & training data

Transforming data to libsvm format

Download MNIST file and extract it in /utils folder:

Downloading & etracting training set

cd utils/
wget https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/multiclass/mnist.t.bz2
wget https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/multiclass/mnist.bz2
bzip2 -d mnist.t.bz2 mnist.bz2

Prepare libsvm files

MNIST files does not contain zero values, which is problematic when using Rumale library. The simple fix is to fill it with zeros, and make sure we have 784 values (28x28 pixels) in one line. For that there is script:

// in utils/
ruby fill_libsvm_with_zeros.rb mnist
ruby fill_libsvm_with_zeros.rb mnist.t

Libsvm files

Libsvm files contain datas in multiple lines with format:

<label> <index1>:<value1> <index2>:<value3> ...
.
.
.

For MNIST data which are just 28x28 pixels images of digits index is the position of pixel (x=index%size, y=floor(index/size)) and value is color of pixel (0-white, 255-black).

Training & testing model

Training model

Now you can start training with train.rb script.

// in utils/
ruby train.rb

Mind that it has parameters such as:

batch_size
max_epoch
validation_split
layers

For training you can tweak them and see how it impacts the learning process and accuracy.

Testing model

To test model simply run:

// in utils/
ruby test.rb

Make sure scipt is using the right model file, and NN has the same dimensions as in training.

Using different model in app

To use you own model copy your .dat and .pth file to models folder, and rename it to model.dat and model.pth. The default model has ~97% accuracy.