# New Release: Ultralytics v8.3.84

**URL:** <https://community.ultralytics.com/t/new-release-ultralytics-v8-3-84/819>\
**Category:** Discussion\
**Tags:** releases, announcements, ultralytics-official\
**Created:** [March 6, 2025, 1:23pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-3-84/819 "2025-03-06T13:23:33Z")\
**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:** [March 6, 2025, 1:23pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-3-84/819/1 "2025-03-06T13:23:33Z")

</div>

# 🚀 Ultralytics v8.3.84 Release Announcement

We are excited to introduce **Ultralytics v8.3.84** , packed with meaningful improvements to segmentation, documentation, and model usability. 🎉 This release focuses on cleaner outputs, easier learning, and enhanced validation capabilities. Read on for all the details!

* * *

## 🌟 Summary

The **v8.3.84** release enhances YOLO’s segmentation accuracy, strengthens documentation, and refines validation settings to provide a more seamless user experience. Highlights include:

- 🌟 **Segmentation Optimization** : Filtering out predictions with empty masks for more focused and usable outputs.
- 📚 **Improved Documentation** : Better instructional materials, including new YouTube resources and clearer code example refinements.
- ⚙ **Validation Refinements** : `save_hybrid` mode is now restricted to detection tasks only, preventing evaluation errors.

* * *

## 📊 New Features & Key Changes

### 🚀 Segmentation Optimization

YOLO models now automatically remove predictions with empty masks, ensuring only meaningful segmentation data is retained. This significantly improves result reliability and clears away irrelevant data for streamlined workflows.

- Implemented in [PR #19537](https://github.com/ultralytics/ultralytics/pull/19537) by [@Y-T-G](https://github.com/Y-T-G).

* * *

### 📚 Updated Documentation

- A **new YouTube guide** for **SAHI Tiled Inference** is now live! Learn more via this [easy-to-follow resource](https://youtu.be/ILqMBah5ZvI). 🎥

- Clearer code examples for tools like `Colors` and `merge_equals_args`, creating a more user-friendly learning path. ✨

* * *

### ⚙ Validation Enhancements

The `save_hybrid` mode has been limited to detection tasks to prevent incorrect validation usage with other models like segmentation or OBB (oriented bounding box). This ensures evaluation results remain accurate across all use cases.

- Details in [PR #19531](https://github.com/ultralytics/ultralytics/pull/19531) by [@Y-T-G](https://github.com/Y-T-G).

* * *

## 🎯 Purpose & Impact

- ✅ **Cleaner Outputs** : Segmentation workflows now offer cleaner, more relevant results by omitting invalid predictions.
- 🤝 **Better Onboarding** : Enhanced documentation ensures a smooth experience for both new and advanced users.
- ⚠ **Safer Validation** : Limiting `save_hybrid` avoids misuse of validation settings, promoting error-free evaluations.

* * *

## 🔗 Changelog & Resources

- View the [full changelog](https://github.com/ultralytics/ultralytics/compare/v8.3.83...v8.3.84) for all updates.
- Access the [v8.3.84 release notes](https://github.com/ultralytics/ultralytics/releases/tag/v8.3.84).

* * *

## 🛠 Try It Out Today!

Update your Ultralytics package to the latest version, explore the improvements, and let us know what you think! Your feedback helps refine our tools to better serve the YOLO community.

```bash
pip install ultralytics --upgrade

```

We encourage you to try out the new features and improvements and share your thoughts on their impact. Drop us a message in the [Ultralytics GitHub Discussions](https://github.com/ultralytics/ultralytics/discussions)—we’d love to hear from you.

* * *

Thank you for being a part of the YOLO journey. All these advancements wouldn’t be possible without the incredible contributions from our community and Ultralytics team! 😊