Ultralytics v8.3.161 Release – Cleaner Dataset Paths, Better Docs, and Enhanced Platform Support!
Summary
Ultralytics v8.3.161 is out! This release focuses on simplifying dataset paths, strengthening documentation, and enhancing cross-platform compatibility—especially on ARM64 devices like Raspberry Pi. You’ll find smoother dataset management, updated guides, clearer deprecation warnings, and a number of performance, reliability, and API improvements.
New Features & Major Improvements
Simplified Dataset Paths
- Dataset YAML files, code samples, and documentation now use plain folder names like
cocoinstead of legacy relative paths like../datasets/coco. - Applies to detection, segmentation, pose, and OBB tasks, making it much easier to specify datasets, especially with custom
datasets_dirsetups. - See details in Eliminate
../datasets/in data yaml files for betterdatasets_dircompatibility by @Laughing-q.
Documentation & Example Updates
- All guides, tutorials, and code snippets are refreshed with the new, cleaner dataset path convention.
- Warning added: Community-contributed examples in
ultralytics/exampleswill be deprecated in v8.4.0. Rely on official docs for the latest support (PR by @Y-T-G).
Improved PaddlePaddle Support
- ARM64 devices now automatically install the compatible
paddlepaddle==3.0.0, enhancing install reliability on platforms like Raspberry Pi (PR by @Laughing-q).
Sony IMX500 Benchmark Updates
- Clarified benchmark results and standardize dataset naming for YOLO11n and YOLOv8n on Sony IMX500 devices (COCO128 Benchmarks & Capitalization Fix by @lakshanthad).
Improvements & Bug Fixes
- VisDrone Label Writing Optimization: Speeds up annotation conversion and makes it more robust (PR by @banu4prasad).
- Verification Accuracy: Corrected expected label count for certain datasets, making verification checks more reliable (PR by @ImJaewooChoi).
- API and Docs Enhancements: New type annotations, improved export and summary methods, and more refined logging (PRs, PR by @RizwanMunawar).
- VisualAISearch Refactor: Streamlined for clarity and consistency (PR by @RizwanMunawar).
Why This Matters
- Easier Dataset Management: Spend less time fiddling with paths—just specify dataset names, and Ultralytics handles the rest.
- Cross-Platform Reliability: ARM64 and Raspberry Pi users benefit from a smoother installation process.
- Onboarding & Maintenance: With updated docs and clearer warnings, it’s easier than ever to get started—and stay up to date.
- Developer Experience: Improved type hints and export documentation make custom workflows and automation more straightforward.
Acknowledgements
A special thank you to new contributor @ImJaewooChoi and to all the community members and Ultralytics team for making this release possible!
Quick Links
Try It Out & Share Feedback!
Upgrade to Ultralytics v8.3.161 to take advantage of these improvements. Your testing, feedback, and contributions are key—let us know what you think right here!
Ready for smoother dataset workflows and cross-platform support? Give this release a spin and share your experience.