# New Release: Ultralytics v8.3.147

**URL:** https://community.ultralytics.com/t/new-release-ultralytics-v8-3-147/1114
**Category:** Discussion
**Tags:** announcements, ultralytics-official, releases
**Created:** [June 2, 2025, 1:13pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-3-147/1114 "2025-06-02T13:13:47Z")
**Posts on this page:** 1
**Page:** 1

<div class="post-metadata">

### Author: ![glenn-jocher](https://sea1.discourse-cdn.com/flex001/user_avatar/community.ultralytics.com/glenn-jocher/32/83_2.png) [@glenn-jocher](https://community.ultralytics.com/u/glenn-jocher)
#### Post date: [June 2, 2025, 1:13pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-3-147/1114/1 "2025-06-02T13:13:47Z")

</div>

# 🚀 Ultralytics v8.3.147 Release: Exportable Confusion Matrices, Enhanced Model Support & Improved Docs!

## 🌟 Summary

Ultralytics v8.3.147 arrives with powerful new ways to export confusion matrix results, upgraded YOLOv7 ONNX/TensorRT inference support, refreshed OpenVINO documentation for YOLO11, as well as several usability and documentation improvements. This release makes analysis easier, expands hardware and model compatibility, and improves the training experience for all users. 📊🚀

* * *

## 📊 New Features

### Export Confusion Matrix Results in Multiple Formats

You can now **export confusion matrix results** after validating your models in CSV, XML, HTML, JSON, and SQL formats. The updated `ConfusionMatrix` class now supports class names and streamlined export functionality, making it easier to analyze, share, and present results in your preferred tools.

- PR: [Confusion Matrix export to CSV, XML, HTML, JSON, and SQL formats by @RizwanMunawar](https://github.com/ultralytics/ultralytics/pull/20834)

* * *

## 🆕 Major Improvements

### YOLOv7 ONNX & TensorRT Inference Support

New guides and scripts help you **export YOLOv7 models to ONNX/TensorRT** and run inference directly with Ultralytics. Please note: only inference (not training) is supported for YOLOv7 in Ultralytics.

- PR: [Add guide for YOLOv7 ONNX inference by @Y-T-G](https://github.com/ultralytics/ultralytics/pull/18453)

### OpenVINO Integration for YOLO11

All OpenVINO documentation now references **YOLO11** and includes updated export instructions, usage examples, benchmarks, hardware compatibility tips, and troubleshooting guidance.

- PR: [OpenVINO docs update by @ambitious-octopus](https://github.com/ultralytics/ultralytics/pull/20731)

* * *

## 📚 Documentation & Usability Updates

- **Prediction Arguments & OBB Documentation**
  - New documentation for the `rect` argument (affecting image padding and speed).
  - Fixed code example for accessing oriented bounding box (OBB) results.
  - PRs: [Add `rect` to prediction arguments by @Y-T-G](https://github.com/ultralytics/ultralytics/pull/20857), [Fix OBB predict example by @Y-T-G](https://github.com/ultralytics/ultralytics/pull/20855)

- **Training Parameter Docstring Update**
  - Training function docstrings now reflect the correct `batch` parameter name.
  - PR: [Update train method docstrings by @erfan-zekri](https://github.com/ultralytics/ultralytics/pull/20888)

* * *

## 🛠 Bug Fixes

- **BatchNorm Initialization Fix**
  - Prevents unintended alterations to BatchNorm statistics when initializing models from YAML files, ensuring more stable and consistent model behavior.
  - PR: [Prevent BatchNorm stats from being changed on model initialization from `yaml` by @Y-T-G](https://github.com/ultralytics/ultralytics/pull/18149)

* * *

## 🎯 Purpose & Impact

- **Flexible, Shareable Analysis:** Export confusion matrices in a variety of formats for use in reports, presentations, or further research.
- **Expanded Model & Hardware Support:** Integrate YOLOv7 ONNX/TensorRT and benefit from the latest OpenVINO optimizations for YOLO11 on Intel hardware.
- **Smoother User Experience:** Clearer documentation and parameter names help avoid confusion for both new and experienced users.
- **Reliable Training:** Model initialization fixes lead to more predictable training, especially with custom architectures.

* * *

## 📝 What’s Changed

- [Add guide for YOLOv7 ONNX inference by @Y-T-G](https://github.com/ultralytics/ultralytics/pull/18453)
- [OpenVINO docs update by @ambitious-octopus](https://github.com/ultralytics/ultralytics/pull/20731)
- [Fix OBB predict example by @Y-T-G](https://github.com/ultralytics/ultralytics/pull/20855)
- [Add `rect` to prediction arguments by @Y-T-G](https://github.com/ultralytics/ultralytics/pull/20857)
- [Prevent BatchNorm stats from being changed on model initialization from `yaml` by @Y-T-G](https://github.com/ultralytics/ultralytics/pull/18149)
- [Update train method docstrings to reflect correct `batch` parameter by @erfan-zekri](https://github.com/ultralytics/ultralytics/pull/20888)
- [Confusion Matrix export enhancements by @RizwanMunawar](https://github.com/ultralytics/ultralytics/pull/20834)

**Full Changelog:** [v8.3.146…v8.3.147](https://github.com/ultralytics/ultralytics/compare/v8.3.146...v8.3.147)

**Release URL:** [Ultralytics v8.3.147 Release](https://github.com/ultralytics/ultralytics/releases/tag/v8.3.147)

* * *

## 🚦 Get Involved

Ready to try out the new features? Upgrade to v8.3.147, explore the new export options and model integrations, and let us know how it works for you! User feedback is essential to keep YOLO evolving—please share your questions, suggestions, and results in the comments.

A huge thank you to every contributor and the entire YOLO community for making these improvements possible. Your feedback and efforts keep the Ultralytics ecosystem vibrant and innovative!

Happy experimenting,  
The Ultralytics Team