📸 A simple application to demonstrate TensorflowJS using mobile net model to predict objects via camera API.
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
config
public
scripts
src
.editorconfig
.gitignore
FindObject.gif
LICENSE
README.md
package.json

README.md

tensorflow-find-object

A simple application to demonstrate tensorflowjs using mobilenet model.

Note: This demo uses the pretrained MobileNet_25_224 model from Keras. It is not trained to recognize human faces. For best performance, show the images of objects like piano, coffee mugs, bottles, etc

Demo

Demo GIF

Installation

yarn or npm install

Run

npm run start

Build

npm run build

This project was bootstrapped with Create React App

Libraries I used

  1. Tensorflow js library.
  2. Used a tensorflow model called mobilenet to predict.
  3. Used media stream API with canvas to capture and give it to mobilenet model.

Problems I faced

  • Got minification error in TensorJS from webpack uglify plugin.
  • Safari browser in iOS device, the media stream API wont work unless you add facingMode = "environment"; in constraints object passed to getUserMedia API + with this hack.
  • After adding to homescreen, media stream API is not working in safari browser in iOS devcie.

MIT Licensed