Tanwar-12/Euro-Coin-Detection

The project implements a custom YOLOv5 detector trained on 1,017 annotated images from Roboflow, using bounding box labeling via label-img and makesense.ai to establish ground truth across a single coin class. Training uses 640×640 image resolution with 300 epochs at batch size 8, with TensorBoard integration for monitoring training metrics and model convergence. The workflow includes dataset preparation with YAML configuration, pre-trained weight initialization, and inference testing on validation imagery to evaluate detection accuracy.

No commits in the last 6 months.