🌟 Summary
Ultralytics YOLO26 has arrived. Re-engineered from the ground up by @glenn-jocher, @Laughing-q, and the Ultralytics YOLO team, YOLO26 is purpose-built for edge and low-power environments. This release introduces a streamlined, native end-to-end NMS-free architecture, delivering faster, lighter, and more accessible deployment across all platforms.
📊 What’s New
- 🚀 NMS-Free End-to-End Inference: A fully native design that eliminates the need for Non-Maximum Suppression (NMS) post-processing, significantly reducing latency and simplifying export complexity.
- ⚡ 43% Faster CPU Inference: Optimized specifically for edge computing, achieving real-time performance on CPU-only devices.
- 🧠 MuSGD Optimizer: A pioneering hybrid optimizer combining SGD with Muon. Inspired by Kimi K2 (Moonshot AI), this brings LLM-grade optimization stability to computer vision training.
- 🔧 Streamlined Architecture (No DFL): Complete removal of Distribution Focal Loss (DFL) to streamline model export and maximize compatibility with low-power hardware.
- 🎯 Task-Specific Enhancements:
- Segmentation: Added semantic loss and multi-scale protos.
- Pose: Implemented RLE for high-precision keypoints.
- OBB: Introduced angle loss to resolve boundary discontinuities.
- 🌐 YOLOE-26 Open-Vocabulary: Zero-shot inference capabilities allow for the detection of any object class using text or visual prompts.
🎯 Impact & Vision
- Edge-First Philosophy: By removing architectural bottlenecks like DFL and NMS, YOLO26 maximizes speed on resource-constrained hardware.
- Unified & Versatile: Full end-to-end support for detection, segmentation, classification, pose, and OBB.
- Next-Gen Training: The MuSGD optimizer bridges the gap between LLMs and Vision, offering faster convergence and stable training.
What's Changed
- Ultralytics Code Refactor https://ultralytics.com/actions by @UltralyticsAssistant in #80
- Update tag.yml by @glenn-jocher in #81
- Add files via upload by @glenn-jocher in #82
- Add files via upload by @glenn-jocher in #83
- OG Images by @sergiuwaxmann in #84
- Ultralytics Refactor https://ultralytics.com/actions by @pderrenger in #86
- Add favicon PNGs by @glenn-jocher in #87
- Update to Tasks Banner by @pderrenger in #88
- Add onboarding assets for application by @RizwanMunawar in #89
- Ultralytics Refactor https://ultralytics.com/actions by @glenn-jocher in #90
- Standardize license headers in TOML/YAML files by @pderrenger in #91
- Compress banners for Ultralytics YOLO11 by @RizwanMunawar in #94
- Update comparison graphic by @glenn-jocher in #96
- Update comparison graphic by @glenn-jocher in #97
- Ultralytics Refactor https://ultralytics.com/actions by @pderrenger in #98
- Add
ultralyticsblog banner by @RizwanMunawar in #99 - Add Permissions to Ultralytics Actions
format.ymlby @glenn-jocher in #100 - Add new logo SVGs for various machine learning frameworks and brands by @onuralpszr in #104
- add TensorFlow logo.js and CoreML SVG files to the project and small name typo fix. by @onuralpszr in #105
- Update Official banners for YV25 by @RizwanMunawar in #106
- Update Official banner for YV25 by @RizwanMunawar in #108
- New YOLO Vision 2025 Shenzhen assets by @glenn-jocher in #109
- Potential fix for code scanning alert no. 1: Workflow does not contain permissions by @glenn-jocher in #110
- Issue templates by @glenn-jocher in #111
New Contributors
- @onuralpszr made their first contribution in #104
Full Changelog: v8.3.0...v8.4.0