Skip to content
All work
AI & MLComputer vision

AI Garbage Detection

A YOLO model that spots and labels 42 kinds of waste in any photo.

Validation photos with the model's bounding boxes around cans, bags, cups and paper

Real model output, 42 waste classes

The problem

Sorting waste by hand is slow and inconsistent. Smart bins and recycling lines need a camera that can tell a glass bottle from a plastic cup on its own.

What I built

  • Trained a custom YOLO26n model on 13,000+ annotated images across 42 waste categories
  • Accelerated training on an Intel Arc GPU using PyTorch XPU
  • Shipped a Streamlit app that draws live bounding boxes, class labels and confidence scores on any uploaded photo

How it works

  1. 1Upload a photo
  2. 2Preprocess the image
  3. 3Run the YOLO detector
  4. 4Draw boxes, labels and confidence
Next projectSentinelAI