Show HN: Fruit Inspector – Real-time YOLO tracking at 40 FPS on old CPUs via Nim

Category: ai-ml

Tags: computer-vision, yolo, nim, onnx-runtime, conveyor-inspection, cpu-inference

Score: 5.0/10 (Innovation: 5, Technical: 6, Documentation: 5, Utility: 4)

Fruit Inspector is a conveyor-belt inspection demo that uses a dual YOLO pipeline (full-frame + ROI refinement) and LightGBM classification to grade apples in a 3x3 control zone, implemented in Nim and ONNX Runtime for CPU-only inference. It achieves real-time performance (~30-115 FPS on an i5 CPU without a GPU), showcasing an interesting edge-computing approach to industrial vision.

Target audience: computer vision engineers, industrial automation developers, embedded AI researchers

Repository: https://github.com/olesha-ai/fruit-inspector-demo · NOASSERTION · 2 stars

View on Hacker News