Ultralytics v8.3.250 is out 

Quick overview: v8.3.250 brings out-of-the-box support for the TT100K traffic sign dataset (auto-download + auto-convert to YOLO labels), plus smoother dataset URI handling, cleaner run directory behavior, and a handful of quality/build/docs fixes ![]()
![]()
New Features
TT100K dataset integration (221 classes)
Big thanks to @PrashantDixit0 for adding first-class TT100K support in PR #22892 ![]()
What you get:
- New dataset config at
ultralytics/cfg/datasets/TT100K.yamlwith full metadata - Auto download + conversion pipeline that fetches TT100K and converts annotations into YOLO-format labels automatically
- New docs page integrated into navigation via the TT100K dataset documentation

Try it:
yolo detect train data=TT100K.yaml model=yolo11n.pt
Improvements
Dataset URI support + automatic NDJSON โ YOLO conversion
PR #23146 by @glenn-jocher improves training workflows when using ul://.../datasets/... sources:
ul://dataset sources are now resolved locally first- NDJSON datasets are automatically converted to YOLO format
- Smarter
--projecthandling for nested names likeuser/project, keeping runs organized under the expected runs directory - Updates to the platform webhook base URL used by the platform callback resolver
Documentation Updates
Better Kaggle install + troubleshooting guidance
PR #22970 by @PrashantDixit0 adds a clearer Kaggle Installation section (internet toggle, install commands, conflict fixes) to reduce setup friction ![]()
Bug Fixes
ConfusionMatrix default type mismatch fix
PR #23138 by @raimbekovm fixes ConfusionMatrix(names=...) default from [] to {} to match the annotated type and avoid mutable-default pitfalls ![]()
Build / Maintenance
ONNXRuntime C++ example build improvement
First contribution from @omar-A-hassan in PR #23137 ![]()
CMake now respects user-provided ONNXRUNTIME_ROOT, making custom install paths work as documented.
Dependency update
PR #23143 by @dependabot[bot] bumps urllib3 in the RT-DETR ONNXRuntime Python example requirements.
How to update
pip install -U ultralytics
Release links
You can review the full release notes in the Ultralytics v8.3.250 GitHub release and browse changes in the full changelog comparison.
Feedback welcome
Give v8.3.250 a spinโespecially TT100K training and ul:// dataset flowsโand reply here with results, issues, or suggestions. Your feedback helps the whole YOLO community and the Ultralytics team keep improving ![]()