Apppublic

trong333tn/ai-quality-control

sourceHugging Faceupdated 5mo agoView on Hugging Face

0likes

App README

Visual Defect Inspector

Real-time surface defect detection system using YOLOv8 + PatchCore. Detects manufacturing defects (scratches, cracks, contamination) from images and video.

Demo

Live demo link (coming soon)

Results

Metric	Value
mAP50	0.783
Precision	0.856
Recall	0.690
Inference speed	~5ms/image (T4 GPU)

Project Structure

visual-defect-inspector/
├── models/          # Model weights (.pt, .onnx)
├── src/
│   ├── detector.py  # Core detection pipeline
│   ├── api.py       # FastAPI REST endpoint
│   └── ui.py        # Gradio demo interface
├── notebooks/       # Training notebooks (Kaggle)
├── data/samples/    # Sample images for demo
├── tests/           # Unit tests
├── Dockerfile
└── requirements.txt

Quick Start

bash

# Install
pip install -r requirements.txt

# Run API
uvicorn src.api:app --reload

# Run UI demo
python src/ui.py

Dataset

—MVTec Anomaly Detection dataset
—15 product categories, 5000+ images
—Training: 1002 defect images
—Validation: 256 defect images