TensorFlow.js Example: Training a baseball model in Node.js

This demo demonstrates how to train a server-side model to classify baseball pitch types using Node.js.

It has four parts:

1. Baseball sensor training and test data.
2. Two ML models that do classification given the sensor data:
   • Model that predicts the type of pitch.
   • Model that predicts if there was a strike.
3. Node.js server that trains a model and serves results over a web socket.
4. Web application that displays pitch type learning statistics.

Running the Demo

First, prepare the environment and download the baseball training and test data:

yarn && yarn download-data

Next, start the client:

yarn start-client

Open the client running at: http://localhost:8080/

In a new shell, start the server:

yarn start-server

If you are interested in testing out the training, without running a web server:

yarn train-pitch-model

yarn train-strike-model

Baseball Models

This demo contains two models. The first is a pitch-type model used in the actual client/server architecture. The other model learns how to call balls and strikes like a major-league umpire. It currently does not have any presentation UI but exists for developers to experiment with.

1. Pitch type model - Classifies 7 different pitch types looking at baseball sensor data (pitch-type-model.ts)
2. Strike zone model - A model that can learns how to call balls and strikes based on historical umpire calls (strike-zone-model.ts).