* [FiftyOne v1.21.0](https://github.com/voxel51/fiftyone) – FiftyOne is a tool for visualizing and labeling datasets, evaluating computer vision models, and improving data and model quality.

* [Scanly v2.0.0](https://github.com/Yangjunjie-Lin/Scanly) – Local-first barcode capture SDK for Web and Node with native mobile packages, featuring multi-format decoding, realtime tracking, batch scanning, and local-only semantic parsing.

* [ppu-paddle-ocr v6.4.0](https://github.com/PT-Perkasa-Pilar-Utama/ppu-paddle-ocr) – Lightweight, TypeScript PaddleOCR-based library for text detection and recognition in Node.js, Bun, and browser environments.

* [React mindee v1.7.0](https://github.com/mindee/react-mindee-js) – Opinionated JavaScript library for building interactive canvases for computer vision detection.

* [react-native-vision-camera-mlkit v2.0.0](https://github.com/pedrol2b/react-native-vision-camera-mlkit) – Nitro-based React Native Vision Camera v5 plugin providing Google ML Kit frame output OCR and barcode scanning for live frames and static images.

* [DSH Vision Toolkit v0.1.19](https://github.com/Anionex/dsh-vision-toolkit) – DeepSeek Harness profile bundle that adds intent-aware image Q&A, OCR, grounding, UI restoration, pixel verification, artifacts, and web settings for text-only agents.

* [VLM-AutoYOLO v1.5.14](https://github.com/Somnusochi/VLM-AutoYOLO) – End-to-end pipeline for VLM-powered image/video auto-annotation, SAM-based mask refinement, manual correction, multi-format export, and one-click YOLO training and validation.

* [Nitro Vision v0.4.0](https://github.com/sagawrr/react-native-nitro-vision-kit) – Nitro Modules-based React Native library for on-device subject cutouts, ranked image labels, and OCR using Apple Vision and Google ML Kit.

* [Cinekive v0.4.4](https://github.com/Gianluca-Improta/cinekive) – Local-first cinematic archive for filmmakers that ingests films, stills, and URLs to search and retrieve frames by look, mood, technique, or director.

* [AI Engineering from Scratch v2026.08](https://github.com/rohitg00/ai-engineering-from-scratch) – A multi-language curriculum that builds AI end-to-end from raw math through agents and production, with reusable artifacts across 20 phases and 503 lessons.

* [light-ocr v0.5.7](https://github.com/arcships/light-ocr) – Fast, offline OCR library for Node.js and C++ that returns recognized text lines with confidence scores and quadrilateral coordinates using hardware acceleration.

* [Facenox v1.0.0-beta.10](https://github.com/facenox/facenox) – Offline-first face recognition software for real-time attendance tracking, designed with local biometric processing and encrypted templates with optional cross-device sync.