BYOE-Vision

Universal visual inspection toolkit — pipeline-first DAG editor with multi-source capture, AI-powered anomaly detection, and headless/edge deployment.

Installation

pip install byoe-vision[all]

Minimal install (core only):

pip install byoe-vision

Optional extras:

• anomaly — Torch-based anomaly detection (autoencoder + PatchCore)
• server — FastAPI inference server + WebSocket streaming
• mqtt — MQTT publish/subscribe nodes
• excel — Excel export support
• dev — Development & testing tools

Quick start

CLI

# Run a pipeline on one image
byoe-vision run pipeline.json image.png

# Process a batch
byoe-vision batch pipeline.json /path/to/images/

# Start the inference server
byoe-vision serve --port 8000

GUI (visual pipeline editor)

byoe-vision-gui

Or directly:

python -m ui.app

What is BYOE-Vision?

Build visual inspection pipelines by wiring together small modules into a DAG:

1. Source — load an image (file, folder, camera)
2. Prepare — crop, fix lighting, threshold
3. Analyze — anomaly detection, blobs, calipers, OCR, template/object detection
4. Decide — turn scores into pass/defect verdicts
5. Show — display results

Modules include: Anomaly Detection (Autoencoder/PatchCore), Object Detection (YOLO), Template Matching, Blob Analysis, Calipers, OCR/Barcode/QR, Fixture alignment, MQTT/HTTP networking, and more.

License

MIT