60 notebooks that actually run: a field guide to modern CV

What it does This repo is a maintained index of 60 Jupyter notebooks covering object detection, segmentation, pose estimation, OCR, and visual-language tasks. Each notebook links to runnable Colab, Kaggle, or SageMaker Studio Lab environments, plus original papers and complementary video walkthroughs. Think of it as a living syllabus for practical computer vision.

The interesting bit The curation is aggressive about freshness: SAM 3, Qwen3-VL, GLM-OCR, RF-DETR segmentation, and YOLO26 all appear alongside older staples like YOLOv8. The table auto-generates from source, so the README doesn’t quietly rot when new models drop.

Key highlights

• 60 model-specific tutorials with one-click cloud execution
• Coverage spans fine-tuning (custom datasets), zero-shot inference, and multi-modal LLM vision
• Each entry maps to original repos, arXiv papers, and often Roboflow blog posts or YouTube videos
• Includes both foundational architectures and week-old releases
• Maintained by a company that actually ships vision infrastructure, not just content

Caveats

• The repo is curation and notebooks, not a framework; you’ll leave with working examples, not a unified API
• Some newer notebooks lack complementary materials (blank columns in the table)
• “SOTA” is the README’s word, not ours; model ages range from months to days

Verdict Great for practitioners who want to kick tires before committing to a model, or for teams onboarding junior CV engineers. Skip if you need a unified training pipeline or deep theoretical exposition.