How vikhram-s/visual-quality-inspector works
This is a full-stack image quality and defect inspection system. A user uploads an image through the React interface; the FastAPI service validates and stores it, invokes feature extraction and ML inference, then returns a score, defect explanations, and a spatial heatmap. Results are persisted for paginated history, individual retrieval, and image access. The repository also includes dataset generation, model training, and synthetic/real-holdout evaluation. Runtime relationships are grounded in the excerpts and README; some model-development wiring is only partially evidenced.
Main entry points: main.py REST API · App.jsx Inspection UI · ml_engine.py ML inference · train.py Model training
Components 15
Inspection interface
- Inspection UI
frontend/src/App.jsx
Analysis service
- REST API
backend/main.py - API schemas
backend/schemas.py
Vision and inference
- ML inference
backend/ml_engine.py - Feature extraction
ml/feature_extractor.py - Spatial heatmap
ml/feature_extractor.py - Inference model
backend/model.joblib
Result persistence
- Database session
backend/database.py - Analysis record
backend/models.py - Analysis database
- Uploaded images
backend/uploads
Model development
- Dataset generation
ml/generate_dataset.py - Model training
ml/train.py - Model evaluation
ml/evaluate.py
Other components
- Inspector
Connections 18
- Inspector → Inspection UI: uploads image
README.md - Inspection UI → REST API: sends REST request
README.md - REST API → API schemas: validates responses
backend/main.py - REST API → ML inference: invokes analysis
backend/main.py - ML inference → Feature extraction: extracts features
backend/ml_engine.py - ML inference → Inference model: loads classifiers
backend/ml_engine.py - REST API → Spatial heatmap: generates heatmap
backend/main.py - REST API → Uploaded images: writes image
backend/main.py - REST API → Database session: gets database session
backend/main.py - Analysis record → Database session: uses ORM base
backend/models.py - REST API → Analysis record: creates and queries records
backend/main.py - Database session → Analysis database: connects to SQLite
README.md - Inspection UI → REST API: requests history
README.md - Dataset generation → Model training: supplies datasets
- Model training → Feature extraction: extracts training features
ml/train.py - Model training → Inference model: saves backend artifact
ml/train.py - Model evaluation → Feature extraction: extracts evaluation features
ml/evaluate.py - Model evaluation → Inference model: evaluates classifiers
ml/evaluate.py