deep-learning-proto

Deep Neural Network for Image Classification

Implements a 2-layer and L-layer neural network based binary classifiers in Scala using linear algebra libraries. There are two different implementations: One based on Breeze and another one with ND4J.

Training data is stored in HDF5 binary data format. NetCDF library is used for loading the data.

Running the web demo

• pre-requisites: nodejs, npm, Java JVM, sbt

• build the front-end

cd front-end
npm install
npm run build

• run backend

cd ..
sbt 'runMain com.practicingtechie.dl.rest.ImageClassificationServer'

Viewing training set images

Images can be viewed with either ImageRendered or renderimage.py.

Using Python image rendering tool

• install Virtualenv
• create new virtualenv and activate it

virtualenv dl-py
source dl-py/bin/activate

• install matplotlib and h5py

pip install matplotlib
pip install h5py