webcam-ml

Train your own real-time image classifier on-the-fly. Real-time image classification using your webcam with tensorflow.js

Features

1. Custom label driven real-time image classification. Learn what anything looks like, and call it what it is! Wide range of applications from medical to meme.
2. Authentication-less model save! Who wants to create an account? Not me... just copy the UUID and save it somewhere!
3. Authentication-less model load! Send the UUID of your trained model to your friends and have them load it! Great for memes.

Demo of App

Usage

1. Focus on an object or entity with your webcam
2. Enter label describing what the webcam is viewing
3. Submit several example captures (ideally with different object angles)
4. Observe the model's classification in real time
5. Repeat to build your own personally trained image classifer
6. Save the model for later! Or send the code to your friends for them to load your model!

Installation for local development

1. Run git clone https://github.com/james-salafatinos/webcam-ml.git in your desired directory
2. Run npm install -i to install the dependencies
3. Run npm start to initiate the server listening at http://localhost:8080/

Installation for Google Cloud deployment

1. Run git clone https://github.com/james-salafatinos/webcam-ml.git in your cloud shell
2. Run npm install -i to install the dependencies
3. Run gcloud app deploy and wait a few minutes to spin up the active server in the cloud

To-dos:

1. [Feature] - Add loading gif for when webcam is initializing and model is downloading
2. [Bug] - Fix unsupported Safari "getUserMedia()"

Tool Landscape

1. Testing Framework
   1. Mocha
2. App Deployment
   1. Google App Engine
3. Editor
   1. VS Code
4. Server
   1. Express
5. Backend
   1. NodeJS
6. Front End
   1. None
7. CI/CD
   1. Github Actions
8. Database
   1. MongoDB

Demo of Saving

Demo of Loading