I always struggle to remember what is recyclable and what isn't so I decided to use Tensorflow.js and the object detection model coco-ssd to recognise objects and help determine which bin I should throw them in.

This is a prototype.

The object detection is only as good as the pre-trained model used.

For the classification between different bins, I wrote a couple of files listing different objects that should be thrown in the yellow or red bins. These objects will differ depending on the country you live in. I based my selection on the NSW region in Australia.

At the moment, the objects listed in my files are based off of the 90 classes of the coco-ssd model and the 1000 classes of the mobilenet model.

The lists could be more detailed and include one for electronic goods that need to be thrown separately but that's a next step.

My main goal so far was to see if it would work, and it kinda does :)

Working Demo

1. Clone the repository:

git clone https://github.com/charliegerard/tfjs-recycling.git

2. Navigate to it:

cd tfjs-recycling

3. Install the dependencies:

yarn

4. Start:

yarn start

A browser window/tab should open on port 8081.

• Tensorflow.js v0.15.3
• Tensorflow model coco-ssd v0.1.1
• HTML / CSS / JS
• Webpack 4
• Babel 7

To build the project, run yarn build and the public directory will have the deployable files.