Waste sorting with object detection

The scenario

BinSight walks through a complete computer-vision workflow: labelling a dataset, training a YOLO detector, evaluating it, and serving it behind an API with a webcam demo. Packaged as a modular project with a training pipeline you can rerun on new data.

System architecture

High-level flow of BinSight, from input to output.

What you will learn

• Dataset labelling and augmentation
• Training and evaluating detectors
• Serving vision models

What you will build

• Training pipeline
• Detection API
• Webcam demo

Build plan

1. Dataset
2. Train
3. Evaluate
4. API
5. Demo

Tech stack