ultralytics/ultralytics v8.3.171

v8.3.171 - `ultralytics 8.3.171` add NVIDIA NGC auth to Docker Action (#21556)

on GitHub

🌟 Summary

This release brings enhanced Docker workflow security with NVIDIA NGC authentication, a comprehensive new guide for deploying YOLO11 on Google Cloud Vertex AI, improved object tracking consistency, and several documentation and usability updates. 🚀🔒

📊 Key Changes

• Secure NVIDIA NGC Access in Docker Builds: Docker GitHub Actions now support secure login to NVIDIA NGC, enabling access to NVIDIA resources during automated builds. 🔒🐳
• Vertex AI Deployment Guide: Added a detailed, step-by-step guide for deploying YOLO11 models on Google Cloud Vertex AI using Docker and FastAPI, making scalable cloud inference more accessible. ☁️📖
• Object Tracking Improvements: Standardized bounding box formats and improved feature handling in BoT-SORT and BYTETracker for more reliable and maintainable tracking. 🏷️🔄
• Cleaner Output Control: Progress bars (TQDM) are now automatically hidden when logging is set above INFO, reducing clutter for advanced users. 🧹
• IMX Export Stability: Pinned the pydantic dependency for IMX model export to prevent compatibility issues, and clarified export file naming in the documentation. 📦📝
• Metrics Documentation: Expanded and unified docstrings for metrics classes, improving clarity and consistency for developers and users. 📚

🎯 Purpose & Impact

• Smoother Docker Automation: Secure NVIDIA NGC authentication ensures Docker builds can access advanced NVIDIA containers and resources, streamlining CI/CD for GPU-based workflows.
• Easier Cloud Deployments: The new Vertex AI guide empowers users to deploy YOLO11 models at scale on Google Cloud, with best practices and troubleshooting tips for production environments.
• More Reliable Tracking: Standardizing tracking code reduces bugs and confusion, making it easier to integrate and extend object tracking features.
• User-Friendly Output: Automatic control of progress bar visibility provides a cleaner experience for users running scripts in production or with custom logging preferences.
• Robust IMX Exports: Dependency pinning and clearer documentation minimize errors and confusion for users exporting models to Sony IMX devices.
• Better Developer Experience: Improved documentation and code comments help both new and experienced users understand and extend the Ultralytics codebase.

Overall, this update strengthens security, usability, and documentation, while making advanced deployment and tracking features more accessible to the community. 🌍✨

What's Changed

• Vertex AI Docker deployment guide for YOLO11 by @lussebullar in #21531
• Fix file naming structure for Sony IMX export by @lakshanthad in #21520
• Update docstrings of functions in metrics.py by @nikunjlad in #21526
• Fix BoT-SORT with ReID when the features are passed natively by @Rusteam in #21509
• Pin pydantic<=2.11.7 by @Laughing-q in #21537
• Disable TQDM output at higher logging levels by @Y-T-G in #21538
• ultralytics 8.3.171 add NVIDIA NGC auth to Docker Action by @glenn-jocher in #21556

New Contributors

• @lussebullar made their first contribution in #21531

Full Changelog: v8.3.170...v8.3.171