All projects
CogniRecycle
Smart City Waste Classification
Waste contamination silently degrades recycling streams. One wrong item in a bin can spoil the whole batch. CogniRecycle watches the stream with two models, cross-checks every classification, and flags contamination as it happens.
Result
Dual-model verification
Multi-class detection
Real-time inference
What it does
Computer-vision waste sorting for smart city pipelines: a YOLO detector classifies waste materials in real time, and a second model cross-checks each result to flag contamination from mixed or incorrect disposal.
What I built
- Dual-model inference pipeline: the primary YOLO model detects and classifies waste, a secondary model reviews each result for contamination, and the two outputs combine into one routing decision.
- Trained on Garbage Classification Dataset 2 with a low-latency pipeline designed for camera and smart-bin integration.
How it works
- 01Camera / Image Input
- 02YOLO Detector (object detection + material classification)
- 03Secondary Model (contamination check)
- 04Post-processing
- 05Class + Contamination Status
Features
- Real-time waste detection and material classification with YOLO (Ultralytics)
- Multi-class recognition across plastic, paper, metal, glass, and other material types
- Secondary contamination model that catches mixed or incorrectly disposed waste
- Combined decision output: material class plus contamination status per item
- Low-latency inference pipeline built for smart bins, sorting systems, and recycling plants