v8.2.53 - `ultralytics 8.2.53` Heatmaps fix for empty images (#14329)

🌟 Summary

Ultralytics v8.2.53 release primarily focuses on updating download links in various scripts, configuration files, and documentation.

📊 Key Changes

• Dockerfiles Update: Changed the source URLs for downloading tensorstore, torch, and torchvision for ARM packages.
• Dataset Configuration: Updated the download URLs for multiple datasets (e.g., ImageNet10, GlobalWheat2020, VOC, VisDrone) to new asset links.
• Documentation Update: Changed image and dataset links in several documentation files.
• Test Scripts Update: Modified test scripts to use new URLs for downloading test resources.
• Minor Code Tweaks: Adjusted the __init__.py version and made small enhancements to heatmap and training scripts.

🎯 Purpose & Impact

• URL Standardization: Streamlining access to resources by consolidating download links to a unified assets repository ensures consistency and reliability.
• Improved User Experience: Simplifies the process for users to find and download necessary components without broken links.
• Enhanced Documentation: Keeping documentation up-to-date with the latest links ensures users can follow instructions without issue.
• Stable Testing Environment: Ensures that test environments remain consistent and dependable with updated resource links.
• Minor, Yet Crucial: These changes, although minor, are crucial for maintaining the usability and stability of the software, especially for automated scripts and pipelines.

What's Changed

• BaseTrainer with find_unused_parameters=True when using DistributedDataParallel() by @Tsai-chia-hsiang in #14323
• Ultralytics Asset URL Update by @glenn-jocher in #14345
• ultralytics 8.2.53 Heatmaps fix for empty images by @ambitious-octopus in #14329

New Contributors

• @Tsai-chia-hsiang made their first contribution in #14323

Full Changelog: v8.2.52...v8.2.53

v8.2.52 - `ultralytics 8.2.52` fix CenterCrop transforms for PIL Image inputs (#14308)

🌟 Summary

Release v8.2.52 of Ultralytics introduces detailed guidance on maintaining computer vision models, along with modifications to various dataset download links and a few minor updates for better functionality.

📊 Key Changes

• New Guide Added: 📘 "Maintaining Your Computer Vision Model."
• Dataset Links Updated: 🗂 Many dataset download links now point to GitHub instead of the Ultralytics website.
• Testing URLs Adjusted: 🔄 Updated certain URLs in test scripts to point to new locations.
• Conversion Update: 🔄 Adjustments to ensure compatibility with torch 1.13+ for some model and dataset handling functions.
• Various Fixes: 🛠️ Including typo corrections, descriptor enhancements, and code improvements.

🎯 Purpose & Impact

• Comprehensive Model Maintenance Guide:

  • Purpose: 📚 To help users understand how to keep their models accurate and functional post-deployment through monitoring, anomaly detection, and retraining strategies.
  • Impact: 🌟 Enhanced reliability and performance of deployed computer vision models; helps users identify and address model drift and anomalies proactively.

• Updated Dataset Links:

  • Purpose: 🌐 Ensure the datasets are accessible with feature improvements using GitHub releases.
  • Impact: 📦 Users will experience streamlined, reliable access to datasets, avoiding potential downtime or slowdowns previously encountered with older links.

• Compatibility Fixes:

  • Purpose: 🔄 To maintain seamless conversion and implementation especially in environments running torch 1.13+.
  • Impact: 🚀 Users working with recent versions of torch will benefit from increased stability and compatibility, facilitating smoother model training and deployment.

In Summary

The v8.2.52 update is geared toward ensuring continuous model effectiveness and improving data access and functional resilience, leading to a more robust and user-friendly experience.

What's Changed

• Allow OpenVINO export from CUDA by @wh1t3h47 in #14256
• Fix end2end attribute in init_criterion by @Alejandro-Casanova in #14267
• Add Maintaining Your Computer Vision Models Docs Page by @abirami-vina in #14304
• allow_empty=True for Classify dataset class directories by @Alexis-IMBERT in #14301
• Fix Annotator PIL Image size (width, height) order by @SheffeyG in #14227
• ultralytics 8.2.52 fix CenterCrop transforms for PIL Image inputs by @glenn-jocher in #14308

New Contributors

• @wh1t3h47 made their first contribution in #14256
• @Alejandro-Casanova made their first contribution in #14267
• @Alexis-IMBERT made their first contribution in #14301
• @SheffeyG made their first contribution in #14227

Full Changelog: v8.2.51...v8.2.52

v8.2.51 - `ultralytics 8.2.51` update YOLOv9 `GITHUB_ASSETS_NAMES` (#14261)

🌟 Summary

Ultralytics v8.2.51 introduces crucial updates for enhanced Docker environments and ensures better TensorRT compatibility, aimed at delivering a more reliable and efficient user experience.

📊 Key Changes

• Dockerfile Enhancements:
  • Addition of wget to the list of installed packages across multiple Dockerfiles.
  • Inclusion of specific version pinning for tensorrt-cu12 to avoid known bugs.
  • Removal of unnecessary TensorRT installation command in the base Dockerfile.
• Documentation Updates:
  • Expanded details on pose estimation models, including an index mapping of keypoints to body parts.
  • Minor tweaks in the Google Colab documentation for better readability.
• Codebase Adjustments:
  • Version pinning changes in TensorRT to ensure compatibility and avoid breaking changes.
  • Inclusion of additional authors in documentation files.

🎯 Purpose & Impact

• Improved Docker Images:
  • 🐳 Adding wget helps in more versatile script execution and package handling inside Docker containers.
  • 🧩 Pinning tensorrt-cu12 to version 10.1.0 prevents compatibility issues and avoids bugs introduced in version 10.2.0.
• Enhanced User Guidance:
  • 📚 Expanded documentation on pose models provides clearer insights for users working on pose estimation tasks, aiding in better understanding and application.
• Increased Stability:
  • 📐 Version-specific requirements for TensorRT guarantee a smoother setup and operation, minimizing unexpected errors during model export and execution.

These updates collectively aim to provide a more robust and user-friendly experience for developers and users working with Ultralytics' tools and models. 🚀

What's Changed

• Pin tensorrt<=10.1.0 to fix libnvinfer_builder_resource_win.so.10.2.0 error by @glenn-jocher in #14239
• Dockerfile install tensorrt-cu12==10.1.0 by @glenn-jocher in #14240
• Update Pose docs with keypoint explanations by @darouwan in #14248
• ultralytics 8.2.51 update YOLOv9 GITHUB_ASSETS_NAMES by @Laughing-q in #14261

Full Changelog: v8.2.50...v8.2.51

v8.2.50 - `ultralytics 8.2.50` new Streamlit live inference Solution (#14210)

🌟 Summary

The v8.2.50 Ultralytics release focuses on documentation updates and adding new features to improve user experience and training efficiency.

📊 Key Changes

• Updated Links:
  • Updated URL in .github/workflows/greetings.yml for better accessibility.
  • Added a badge for Ultralytics Forums in README.md files across different translations.
• Documentation Enhancements:
  • Added and improved FAQs sections for various datasets.
  • Included detailed instructions and examples in the dataset documentation for training YOLO models.
  • Improved formatting in docs/build_docs.py for better readability.
• New YAML configuration files:
  • Added detailed YAML configuration files for various datasets.

🎯 Purpose & Impact

• Improved User Experience:
  • 📝 Enhanced Documentation: The updated FAQs, examples, and improved formatting make it easier for users to understand and utilize the datasets for training YOLO models.
  • 🔗 Better Accessibility: Updated URL links ensure that users can access the most relevant and up-to-date information with ease.
• Community Engagement:
  • 💬 Forums Badge: Integration with Ultralytics Forums encourages community engagement and provides a platform for user interactions and support.
• Training Efficiency:
  • 📊 Configuration Files: The new YAML configurations simplify dataset setup for training YOLO models, making it straightforward for both novice and expert users.

What's Changed

• Add FAQs to Docs Datasets and Help sections by @glenn-jocher in #14211
• Fix mkdocs.yml raw image URLs by @glenn-jocher in #14213
• Fix Action Recognition Example with torch>=2.0 by @fcakyon in #14232
• Add Discourse at https://community.ultralytics.com by @glenn-jocher in #14231
• ultralytics 8.2.50 new Streamlit live inference Solution by @glenn-jocher in #14210

Full Changelog: v8.2.49...v8.2.50

v8.2.49 - `ultralytics 8.2.49` fix classification `setup_model` (#14199)

🌟 Summary

Ultralytics v8.2.49 introduces new documentation and guides for deploying YOLO models on NVIDIA Jetson devices using DeepStream SDK and TensorRT, along with enhancements to existing documentation.

📊 Key Changes

• New Guide: Added a detailed guide for deploying YOLOv8 on NVIDIA Jetson devices using DeepStream SDK and TensorRT.
• Documentation Update: Included additional guides on model deployment practices and model testing.
• Improved Index: Enhanced the 'Guides' index to include the latest documentation updates.

🎯 Purpose & Impact

• Enhanced Capabilities: The new guide empowers users to deploy YOLOv8 models on Jetson devices, maximizing inference performance using NVIDIA's DeepStream SDK and TensorRT. 📈
• Broadened Knowledge Base: Additional guides on best practices for model deployment and testing provide comprehensive insights to improve model performance and reliability. 📚
• User-Friendly: Improved documentation structure makes it easier for users, both new and experienced, to find and utilize resources effectively. 🗂️

What's Changed

• Default strip_optimizer() to use_dill=False by @glenn-jocher in #14107
• Ultralytics Code Refactor https://ultralytics.com/actions by @glenn-jocher in #14109
• Update NVIDIA Jetson DeepStream Guide with YOLOv8 and Jetson Orin Support by @lakshanthad in #14059
• Update FAQ.md by @glenn-jocher in #14134
• Update Results and CFG docstrings by @glenn-jocher in #14139
• Add Docs models pages FAQs by @glenn-jocher in #14167
• Add Model Testing Guide and Best Practices for Model Deployment Guide by @abirami-vina in #14105
• Add https://youtu.be/mUybgOlSxxA to docs by @RizwanMunawar in #14195
• Add FAQ sections to Modes, Tasks, Usage by @glenn-jocher in #14181
• ultralytics 8.2.49 fix classification setup_model by @Laughing-q in #14199

Full Changelog: v8.2.48...v8.2.49

v8.2.48 - `ultralytics 8.2.48` strip model `criterion` on save (#14106)

🌟 Summary

A minor update with key improvements and refinements to model loading and processing.

📊 Key Changes

• Improved Error Messaging: Updated error and warning messages to generalize references from "YOLOv8" models to "Ultralytics" models.
• Model Verification: Enhanced checks to ensure files are valid Ultralytics models before processing.
• Stripping Optimizer Function: Added and refined steps within the strip_optimizer function:
  • Replaced model with Exponential Moving Average (EMA) if available.
  • Stripped unnecessary components (optimizer, loss criterion, etc.) to reduce file size.
  • Converted specific attributes to ensure compatibility and effective usage.

🎯 Purpose & Impact

• Improved Clarity: The changes in messaging make it clearer for users regardless of which version of the model they are using, reducing potential confusion.
• Efficient Model Handling: By verifying model validity and stripping non-essential parts, the update ensures smoother operations and smaller file sizes, potentially leading to faster loading times and less storage usage.
• Enhanced Flexibility: The refined error and warning messages allow for a more robust handling of various model scenarios, increasing the flexibility and robustness of the software for users.

What's Changed

• ultralytics 8.2.48 strip model criterion on save by @glenn-jocher in #14106

Full Changelog: v8.2.47...v8.2.48

v8.2.47 - `ultralytics 8.2.47` YOLOv8 zero-shot action recognition example (#13935)

🌟 Summary

Ultralytics v8.2.47 introduces new features and enhancements, mainly focusing on documentation updates, action recognition examples, and minor code improvements.

📊 Key Changes

• Documentation Enhancements:
  • Added detailed sections on Fashion-MNIST dataset, highlighting its usage with a video tutorial embed.
  • Introduced a new guide on Model Evaluation and Fine-Tuning.
  • Updated the AI Gym workout monitoring guide.
  • Improved loss function documentation.
• New Examples:
  • Added a comprehensive example for Action Recognition using YOLOv8, including an in-depth guide and scripts for real-time video action recognition.
• Code Improvements:
  • Renamed internal configurations to follow the 'yolov10' naming convention.
  • Simplified loss computation classes and functions.
  • General improvements to better handle variable image sizes and detailed internal metric extraction in YOLOv8.

🎯 Purpose & Impact

• Documentation Enhancements:
  • 📚 Provides users with more comprehensive guides and tutorials for better understanding and implementing various features in Ultralytics.
  • 🎦 The Fashion-MNIST video tutorial makes it easier for newcomers to start with image classification tasks.
  • 🛠 The new guide on model evaluation and fine-tuning helps users optimize their models more effectively, improving overall model performance.
• New Examples:
  • 🎥 The action recognition example enables users to leverage zero-shot video classification, expanding the range of applications for YOLOv8, particularly in video surveillance and behavioral analysis.
• Code Improvements:
  • 🧹 Cleans up and organizes internal configurations, making it easier for developers to navigate and understand the codebase.
  • 🚀 Simplifies the loss computation process, which could lead to more efficient and readable loss calculation workflows.
  • 🔧 Ensures better handling of varied input image sizes, making YOLOv8 more versatile for different datasets and use cases.

What's Changed

• Add https://youtu.be/eX5ad6udQ9Q to docs by @RizwanMunawar in #14077
• Fix deprecation warning by @glenn-jocher in #14091
• Add Insights on Model Evaluation and Fine-Tuning Docs Page by @abirami-vina in #14085
• Strip dfl_loss from BboxLoss by @Laughing-q in #14041
• Results, DFL and AIGym fixes by @ambitious-octopus in #14074
• ultralytics 8.2.47 YOLOv8 zero-shot action recognition example by @fcakyon in #13935

Full Changelog: v8.2.46...v8.2.47

v8.2.46 - `ultralytics 8.2.46` fix OBB Results `xyxy` attribute (#14020)

🌟 Summary

Ultralytics has released version 8.2.46 with updates aimed at enhancing compatibility and usability.

📊 Key Changes

• Dependency Adjustment: Modified the numpy version requirement from ">=1.23.5,<2.0.0" to ">=1.23.0,<2.0.0".
• Enhanced Logging: Updated numpy attributes logging in tests to print before and after result methods.
• Optimized Code: Refactored the xyxy method in results handling for more efficient computations.

🎯 Purpose & Impact

• Dependency Flexibility: 🛠️ The adjustment in numpy dependency broadens compatibility by allowing a wider range of numpy versions. This could help users avoid conflicts with other projects dependent on different versions of numpy.
• Improved Debugging: 🐞 Enhanced logging in tests helps developers by providing clearer insights into the numpy attributes, simplifying debugging and validation processes.
• Performance & Readability: ⚡ The optimized xyxy method improves performance and maintainability by utilizing more efficient computations for bounding box transformations, benefiting both expert developers and applications that require fast, accurate results.

These updates contribute to a more robust and user-friendly experience for developers working with the Ultralytics YOLO framework. 🚀

What's Changed

• ultralytics 8.2.46 fix OBB Results xyxy attribute by @glenn-jocher in #14020

Full Changelog: v8.2.45...v8.2.46

v8.2.45 - `ultralytics 8.2.45` Fix YOLOv8 `augment` inference (#14017)

🌟 Summary

Ultralytics v8.2.45 enhances model training guidance and prediction functionalities.

📊 Key Changes

• Documentation Enhancements:

  • Improved explanations in model training tips including batch size adjustment, early stopping, and training environments.
  • Consistent use of contractions for a more conversational tone.

• Code Updates:

  • Updated model prediction test to include an augmentation parameter.
  • Small enhancement to internal model prediction function to check for 'end2end' attribute.

• Contributor Update:

  • Added a new contributor to the repository (zhixuwei).

🎯 Purpose & Impact

• Documentation Clarity 💡:

  • Purpose: Simplifies complex training concepts for better user understanding.
  • Impact: Users can more easily optimize their model training, leading to better model performance and resource utilization.

• Enhanced Testing ✅:

  • Purpose: Ensures robustness in handling batch predictions with image augmentations.
  • Impact: Enhances the reliability of prediction outputs, providing users with more dependable results.

• Inclusivity & Acknowledgment 🌐:

  • Purpose: Update contributor list, recognizing ongoing community support.
  • Impact: Encourages more community contributions by fostering a welcoming environment.

What's Changed

• ultralytics 8.2.45 Fix YOLOv8 augment inference by @Laughing-q in #14017

Full Changelog: v8.2.44...v8.2.45

v8.2.44 - `ultralytics 8.2.44` Increase Predict dataloader robustness (#14005)

🌟 Summary

The v8.2.44 release focuses on enhancing model training documentation, improving data loader error handling, and adding new model training tips.

📊 Key Changes

• GitHub Actions: Simplified workflow triggers by removing unnecessary push events.
• Documentation Updates:
  • Added a new guide titled "Tips for Model Training" with extensive best practices and optimization strategies.
  • Enhanced various sections by including more detailed instructions and formatting adjustments.
• Source Code Improvements:
  • Improved image read error handling in data loaders by changing from raising errors to logging warnings.
  • Added detailed docstrings and typing annotations in the plotting methods for better code readability and maintenance.

🎯 Purpose & Impact

• Improved Documentation:
  • 🎓 The new model training tips guide provides comprehensive strategies for efficient training, benefiting both beginners and experienced users.
  • 📚 Detailed and structured documentation enhances learning and application of YOLO models, helping users achieve better results.
• Better Error Handling:
  • 🚀 Improved error handling in data loaders ensures smooth execution, making the system more robust against corrupt or missing image files.
  • 🔍 Logging warnings instead of raising errors allows the training process to continue, minimizing interruptions and saving users’ time.
• Enhanced Code Quality:
  • 🛠️ Improved documentations and added typing in the source code, leading to easier maintenance and better developer experience.

These updates collectively aim to streamline the user experience, boost model performance, and ensure more robust and reliable training workflows.

What's Changed

• Added a max_size parameter to the plot_images function by @bobyard-com in #14002
• Add Tips for Model Training Docs Page by @abirami-vina in #14011
• ultralytics 8.2.44 Increase Predict dataloader robustness by @zhixuwei in #14005

New Contributors

• @zhixuwei made their first contribution in #14005

Full Changelog: v8.2.43...v8.2.44