Packages, developer tools, research code, and applications I maintain or created.
- SAHI (5,530 stars): Sliced inference for small object detection.
- Claude & Codex Settings (1,161 stars): Agent skills, hooks, and plugin marketplaces.
- PhD Skills (413 stars): Research workflows for paper reproduction and review.
- Ultralytics (62,213 stars): World's most used real-time vision AI models across 7 tasks.
- Claude Code with Codex (19 stars): A Rust package for using Claude Code with Claude, Codex, Kimi, Grok, and Cursor.
- Image Comparison (254 stars): A responsive before-and-after image component.
- SafeText (57 stars): Profanity detection and filtering for 13 languages.
- Moderators (7 stars): Run content moderation models through one Python API or CLI.
- Balanced Loss (98 stars): Class-balanced cross-entropy and focal loss for PyTorch.
- LabelMe to COCO (321 stars): Convert LabelMe annotations to COCO format.
- MIDV-500 (106 stars): Download MIDV-500 and convert its annotations to COCO.
- YOLOv5 Pip (293 stars): A packaged YOLOv5 distribution with added utilities.
- Ultralytics Plus (87 stars): Extra integrations for Ultralytics models.
- Video Transformers (150 stars): Transformer workflows for video understanding.
- CRAFT Text Detector (276 stars): A packaged, cross-platform CRAFT text detector.
- AutoLLM (1,005 stars): Utilities for building applications with large language models.
- Content Moderation (406 stars): Multimodal deep-learning tools for content moderation.
- Small Object Detection Benchmark (202 stars): ICIP 2022 experiments for sliced object detection.
- SenBen (7 stars): The official benchmark and code for explainable content moderation.
- Ultralytics Lite (31 stars): A fast, local workspace for coding agents and your shell.