New Release: Ultralytics v8.3.21

:rocket: Announcing Ultralytics v8.3.21 Release!

:star2: Summary

We’re thrilled to announce the release of Ultralytics v8.3.21! This update brings NVIDIA Deep Learning Accelerator (DLA) export support, along with several documentation and usability enhancements. Dive into the details below and explore how these improvements can benefit your projects.

:bar_chart: Key Changes

NVIDIA DLA Support

  • Export Options for NVIDIA DLA: Now you can export models for NVIDIA DLA on Jetson devices, enabling more energy-efficient inference. Perfect for applications where power consumption is a priority.

Documentation Updates

  • Enhanced Guides: We’ve clarified the guides on using TensorRT and DLA for Jetson devices and updated image URLs for consistency.

Comet Integration

  • Improved Logging: Enhanced logging of plots and metrics for better tracking during model training and evaluation.

Parameter Additions

  • New Parameters: Introduced project and name parameters to help organize prediction and validation outputs more effectively.

Dataset Naming Standardization

  • Consistent Naming: Changed “Roboflow 100” to “RF100” for precision in documentation.

:dart: Purpose & Impact

  • Energy Efficiency: Leverage NVIDIA DLA to reduce power consumption during inference, ideal for energy-conscious applications. :zap:
  • Enhanced User Experience: Updated documentation and new parameters make managing results easier and more intuitive. :blue_book:
  • Improved Metric Tracking: Expanded Comet integration supports comprehensive model evaluation. :bar_chart:
  • Consistency and Reliability: Documentation improvements ensure up-to-date links and names, reducing confusion. :link:

What’s Changed

New Contributors

Full Changelog: v8.3.20…v8.3.21

We invite you to try out the new release and share your feedback with us. Your insights are invaluable in helping us improve and innovate. Happy experimenting!

Release URL