ILearnMachineLearning.js

Description

Thanks to this project I could combine my knowledge from two areas: machine learning and web development. The whole application is written in React using TypeScript and Redux, but the engine that drives it is TensorFlow.js - a modern library for training and deploying machine learning models. You can find a working app here or play with it on your computer.

# install dependencies
npm install

# serve with hot reload at localhost:3000
npm start

MNIST

MNIST dataset is widely known in the field of machine learning. It contains about 70000 examples of handwritten digits. Photos are black and white, normalized to fit into a 28x28 pixel bounding box and anti-aliased. This iconic dataset was used to train the model in Keras and then turned into a form understandable to TensorFlow.js and used as the heart of the first project. The whole process of creating the model has been described in the Medium article.

YOLO

Originally written in Darknet - open source neural network framework  -  YOLO performs really well in the tasks of locating and recognizing objects on the pictures. After converting to a form understandable to TensorFlow.js, I used Tiny YOLOv2 model to recognize objects in pictures sent by the user. All predictions are made on the client's side, without the need to send photos to the server.

Build Keras models

If you are interested in neural networks and would like to know how to create models used in this project, I encourage you to test the scripts on your computer.

# clone repository
git clone https://github.com/SkalskiP/ILearnMachineLearning.js.git

# navigate to models directory
cd ILearnMachineLearning.js/models

# install all required packages and dependencies as well as python environment
./setup.sh

# activate python environment
source .env/bin/activate

Related projects

If you're interested in machine learning, check out my other projects and check my profile on Kaggle.

Read more about In-Browser ML

1. TensorFlow.js Tutorials & Guides
2. Introducing TensorFlow.js: Machine Learning in Javascript
3. Neural Network 3D Simulation
4. YOLO: Real-Time Object Detection
5. AI Experiments with Google
6. Gentle guide on how YOLO Object Localization works with Keras
7. How to implement a YOLO (v3) object detector from scratch in PyTorch: Part 1
8. You only look once