Ultralytics v8.3.146: Grayscale Detection, COCO8-Grayscale, Model Updates, and More!
Summary
We’re excited to announce Ultralytics v8.3.146, a release focused on empowering grayscale object detection workflows! The headline features include the all-new COCO8-Grayscale dataset, the dedicated grayscale yolo11n-grayscale.pt model, and major documentation and ecosystem enhancements to streamline your computer vision projects — whether you’re working with color or grayscale imagery.
New Features
COCO8-Grayscale Dataset
- What’s new:
A small, single-channel version of COCO8 is now available for lightning-fast testing and debugging. - What’s inside:
Includes a ready-to-use YAML config, a dataset download script, and in-depth COCO8-Grayscale documentation. - Who contributed:
See PR #20827 by @Laughing-q
Grayscale Model Support
- Model:
A dedicated yolo11n-grayscale.pt model (single-channel input trained and tested on grayscale data). - Pipeline:
Training, validation, and prediction workflows now fully support grayscale scenarios — tested and documented.
Documentation & Ecosystem Improvements
- COCO8-Grayscale Docs:
New COCO8-Grayscale dataset page with examples, FAQs, and Ultralytics HUB integration guidance. - Dataset Indexes Updated:
Easy discovery and use within all Ultralytics workflows. - Code Quality Upgrades:
- Improved type hints and docstrings (#20812 by @Laughing-q, #20824 by @RizwanMunawar)
- Cleanups for style, readability, and reliability (#20803, #20829 by @RizwanMunawar, #20829)
Other Improvements
- Deprecation Notices:
Added a clear deprecation warning for Neural Magic integrations (#20817 by @Burhan-Q) - Performance:
Analytics chart rendering is now faster and looks better (#20823 by @RizwanMunawar) - Tracker & CLI:
Minor but impactful fixes for ReID CLI workflow (#20814 by @Y-T-G), duplicate warning removed from tracker docs (#20826 by @Y-T-G) - Benchmarking:
Refactored to use format names, boosting clarity and usability (#20819 by @lakshanthad) - Validation:
Removed redundantsave_cropfrom validation arguments (#20821 by @Y-T-G) - Installer Fix:
uvnow respects environment context (#20837 by @Burhan-Q)
Why This Matters
- Unlock Grayscale R&D:
Experiment with medical imaging, industrial inspection, OCR, and other fields where grayscale or single-channel imagery is standard. - Prototype Rapidly:
COCO8-Grayscale accelerates workflow validation and lets you debug pipelines before scaling up. - Enjoy a Smoother Experience:
Improved analytics, cleaner docs, and robust model handling make Ultralytics more reliable and pleasant for everyone. - Stay Ahead:
Deprecation notices help avoid outdated integrations; our continual improvements keep you future-ready.
Links & Changelog
Get Involved!
We invite you to update to v8.3.146, try the new grayscale workflows, and share your feedback or questions with the team and the community. Your input drives our innovation and helps the platform grow stronger!
If you spot a bug or have a feature request, just let us know — together, thanks to our amazing community and contributors, YOLO keeps getting better.
Happy experimenting!
— The Ultralytics Team