FH_2023_TECS: Fully Green

Recycling Recognition Application

Target audience:

• Anyone trying to reduce waste/contribute towards recycling

Goal:

• Educate people on and assist with recycling; Reduce waste and facilitate recycling

Future implementations:

• Maps API to search for nearby recycling facilities
• More classified data to improve model accuracy
• Crowdsourcing feature: allow users to submit data about recyclable materials
• Recycling tracker (gamified recycling, i.e. “You recycled 100 items this month!...”)

Features:

• Recycling info page (i.e. built-in wiki with information regarding recyclable materials)
  • Fragment with image buttons and dialog fragments
• Image recognition for recyclable items
  • Uses built-in camera API; sends image data to an image-recognition AI to verify if the image contains a recyclable item or not
  • Spent time training AI using 22500-image dataset of organic/recyclable materials
• Manual recycling checker
  • User manually inputs information about the item and the data is queried in a database, returning relevant information upon success
  • Uses a text similarity model based on Euclidean distance
  • Trained on 83-count text-based materials classified by recyclable and nonrecyclabe

Tools

• Android Studio
• Back4app (Parse)
• Jupyter Notebook

fullygreen2.mp4