/EarthEyes

♻️ A webapp to recognize objects and tell you where they can be recycled. Repository for MetroHacks, May 25-26, 2019, for Mingle Li and Jet Chung.

Primary LanguageJavaScript

EarthEyes

SECOND PLACE | ENVIRONMENTAL

A Progressive Web App with computer vision and API access to recognize objects and tell you where they can be recycled (+ more!).

How does it work?

Open, click camera button, wait, bam! Hit navigation icon for directions to the recycling center.

  • Swipe left to access manual search and quick-recycling links
  • Swipe right to access (future) cool stuff!

YouTube Video: https://youtu.be/RS1ekTo6DqM

Technologies

  • PWA with Bootstrap, LESS/Flexbox, and jQuery
  • Node.js for server-side controller
  • Tensorflow.js with MobileNet model for object detection and recognition
  • Earth911 API for data on materials and recycling centers
  • Google App Engine to host Node server over HTTPS

Issues

  • Model is limited in detection and speed. Solution: train own model with Python and Keras.
  • API requests result in 500. Solution: Switch to another hosting platform; GAE blocks some outgoing requests for god-knows-why.

Stats:

Main color: #026928

Development time: 17 hours

Cups of coffee: 5

Hours of sleep: 3

Presentation time given: 3 min

Percent of points explained: 45%

Amount of money on Starbucks Gift Card prize when it said $15-$500: $15