# New Release: Ultralytics v8.4.175

**URL:** <https://community.ultralytics.com/t/new-release-ultralytics-v8-4-175/2257>\
**Category:** Discussion\
**Tags:** ultralytics-official, releases, announcements\
**Created:** [October 10, 2026, 1:09pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-4-175/2257 "2026-10-10T13:09:24Z")\
**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:** [October 10, 2026, 1:09pm UTC](https://community.ultralytics.com/t/new-release-ultralytics-v8-4-175/2257/1 "2026-10-10T13:09:24Z")

</div>

# 🚀 Ultralytics v8.4.175: More Reliable Training, Validation, and Deployment

**Ultralytics YOLO v8.4.175 is focused on reliability across existing models—no new model architecture this time.** The headline improvement prevents classification images from remaining in both training and validation splits after you change the split ratio, helping keep evaluation results trustworthy.

This release also improves pretrained weight loading, validation coverage, image and stream handling, export compatibility, tracking, and depth prediction. Explore the [v8.4.175 release notes](https://github.com/ultralytics/ultralytics/releases/tag/v8.4.175) for the complete release overview.

## 🗂 Training and Dataset Improvements

Changing your classification split ratio should not introduce train/validation leakage. Reassigned images now have their old split copies removed, while nested image paths and existing cache and split directories are preserved.

- **Safer classification workflows:** [PR #26597](https://github.com/ultralytics/ultralytics/pull/26597) by [@Vaishnavi220506](https://github.com/Vaishnavi220506) improves classification training and validation, including split cleanup.
- **Nested classification images:** [PR #26598](https://github.com/ultralytics/ultralytics/pull/26598) by [@Nikhi00718](https://github.com/Nikhi00718) preserves nested images during automatic splitting.
- **Better pretrained weight transfer:** [PR #26570](https://github.com/ultralytics/ultralytics/pull/26570) by [@raimbekovm](https://github.com/raimbekovm) transfers class names when appropriate and adapts RGB first-convolution weights for grayscale models.
- **More effective fine-tuning:** [PR #26589](https://github.com/ultralytics/ultralytics/pull/26589) by [@fcakyon](https://github.com/fcakyon) retains class-branch features when fine-tuning to fewer classes.
- **Pose metadata:** [PR #26574](https://github.com/ultralytics/ultralytics/pull/26574) by [@zkoymen](https://github.com/zkoymen) sets `kpt_shape` for pose models built from YAML.
- **Class-weight handling:** [PR #26599](https://github.com/ultralytics/ultralytics/pull/26599) by [@raimbekovm](https://github.com/raimbekovm) keeps classes without training labels neutral in `cls_pw` weights.
- **Text-prompt training compatibility:** [PR #26600](https://github.com/ultralytics/ultralytics/pull/26600) by [@raimbekovm](https://github.com/raimbekovm) skips `cls_pw` in World and YOLOE text-prompt trainers.

## 📈 More Trustworthy Validation

These fixes help ensure validation includes the expected samples and handles annotation and mask edge cases correctly.

- **Complete validation batches:** [PR #26565](https://github.com/ultralytics/ultralytics/pull/26565) by [@raimbekovm](https://github.com/raimbekovm) fixes `drop_last` behavior for compiled validation and distributed training shards.
- **Large semantic-segmentation counts:** [PR #26561](https://github.com/ultralytics/ultralytics/pull/26561) by [@aswanth-07](https://github.com/aswanth-07) uses `int64` pixel counts in semantic validation.
- **COCO keypoint annotations:** [PR #26564](https://github.com/ultralytics/ultralytics/pull/26564) by [@8rulerstar](https://github.com/8rulerstar) skips annotations without labeled keypoints during conversion.
- **Semantic data handling:** [PR #26566](https://github.com/ultralytics/ultralytics/pull/26566) by [@raimbekovm](https://github.com/raimbekovm) fixes semantic mask reading, class-count caching, and box labels.
- **Mask alignment:** [PR #26607](https://github.com/ultralytics/ultralytics/pull/26607) by [@raimbekovm](https://github.com/raimbekovm) crops letterbox padding on whole pixels for retina and SAM masks.

## 🖼 Image, Stream, Tracking, and Depth Fixes

- **16-bit image inputs:** [PR #26577](https://github.com/ultralytics/ultralytics/pull/26577) by [@raimbekovm](https://github.com/raimbekovm) converts 16-bit TIFF and PIL images to 8-bit before inference.
- **Image URL and stream routing:** [PR #26568](https://github.com/ultralytics/ultralytics/pull/26568) by [@raimbekovm](https://github.com/raimbekovm) correctly routes extensionless image URLs and RTSP/RTMP/TCP streams with video-like suffixes.
- **RGBA TIFF compatibility:** [PR #26611](https://github.com/ultralytics/ultralytics/pull/26611) by [@raimbekovm](https://github.com/raimbekovm) reads RGBA TIFFs as three channels for three-channel models.
- **Inference and cache robustness:** [PR #26579](https://github.com/ultralytics/ultralytics/pull/26579) by [@cainiao33](https://github.com/cainiao33) fixes stale NumPy cache reads, multi-stream tracking, custom predictors, CLI brackets, and CPU FP16 handling.
- **Additional pipeline fixes:** [PR #26604](https://github.com/ultralytics/ultralytics/pull/26604) by [@cainiao33](https://github.com/cainiao33) addresses image caching, export batching, DOTA merging, and segmentation heads.
- **Independent tracker IDs:** [PR #26591](https://github.com/ultralytics/ultralytics/pull/26591) by [@raimbekovm](https://github.com/raimbekovm) gives each tracker its own track-ID counter.
- **Stable ReID batch outputs:** [PR #26610](https://github.com/ultralytics/ultralytics/pull/26610) by [@raimbekovm](https://github.com/raimbekovm) clones each fixed-batch ReID chunk output before the next inference call.
- **Rectangular ReID inputs:** [PR #26596](https://github.com/ultralytics/ultralytics/pull/26596) by [@aswanth-07](https://github.com/aswanth-07) fixes rectangular ONNX inputs in tracker ReID preprocessing.
- **Consistent depth preprocessing:** [PR #26556](https://github.com/ultralytics/ultralytics/pull/26556) by [@Y-T-G](https://github.com/Y-T-G) aligns depth prediction image stretching with validation and adds calibration guidance without a depth camera.

## 📦 Export and Runtime Compatibility

- **Embedded NMS and CoreML:** [PR #26585](https://github.com/ultralytics/ultralytics/pull/26585) by [@raimbekovm](https://github.com/raimbekovm) fixes larger inference batches for embedded-NMS exports and empty-frame handling in CoreML.
- **Dynamic YOLOE exports:** [PR #26567](https://github.com/ultralytics/ultralytics/pull/26567) by [@Y-T-G](https://github.com/Y-T-G) fixes YOLOE CoreML export with `dynamic=True`.
- **Static TorchScript validation:** [PR #26554](https://github.com/ultralytics/ultralytics/pull/26554) by [@amanharshx](https://github.com/amanharshx) validates static TorchScript exports at their exported image size.
- **Complete ONNX backend outputs:** [PR #26606](https://github.com/ultralytics/ultralytics/pull/26606) by [@raimbekovm](https://github.com/raimbekovm) returns every output from the OpenCV DNN ONNX backend.
- **Dependency requirement checks:** [PR #26575](https://github.com/ultralytics/ultralytics/pull/26575) by [@shoutoutuoadi325](https://github.com/shoutoutuoadi325) enforces version specifiers for extras and `~=` requirements.

## 📚 Documentation and Tooling Updates

- **Security guidance:** [PR #26552](https://github.com/ultralytics/ultralytics/pull/26552) by [@glenn-jocher](https://github.com/glenn-jocher) expands the security help page.
- **Agent Skills documentation:** [PR #26558](https://github.com/ultralytics/ultralytics/pull/26558) by [@glenn-jocher](https://github.com/glenn-jocher) expands the Agent Skills guide.
- **Unified documentation diagrams:** [PR #26573](https://github.com/ultralytics/ultralytics/pull/26573) by [@glenn-jocher](https://github.com/glenn-jocher) migrates Mermaid diagrams to the unified design system.
- **Clearer Platform workflows:** [PR #26562](https://github.com/ultralytics/ultralytics/pull/26562) by [@raimbekovm](https://github.com/raimbekovm) aligns diagrams with current Ultralytics Platform behavior and corrects Axelera labels.
- **Dataset discovery and metadata:** [PR #26588](https://github.com/ultralytics/ultralytics/pull/26588) by [@raimbekovm](https://github.com/raimbekovm) documents Explore filters and dataset metadata in version comparisons.
- **Updated Platform FAQs:** [PR #26586](https://github.com/ultralytics/ultralytics/pull/26586) by [@miles-deans-ultralytics](https://github.com/miles-deans-ultralytics) refreshes common Platform questions and answers.
- **Rectangular validation guidance:** [PR #26594](https://github.com/ultralytics/ultralytics/pull/26594) by [@maisternia](https://github.com/maisternia) documents the half-stride border used in rectangular validation.
- **Edge-device benchmarks:** [PR #26614](https://github.com/ultralytics/ultralytics/pull/26614) by [@lakshanthad](https://github.com/lakshanthad) adds Raspberry Pi 5 + AI HAT+ Hailo-8 benchmarks.
- **Simpler CI environment setup:** [PR #26557](https://github.com/ultralytics/ultralytics/pull/26557) by [@glenn-jocher](https://github.com/glenn-jocher) uses `setup-uv` defaults for Python and the virtual environment.
- **AI-assisted CI failure analysis:** [PR #26553](https://github.com/ultralytics/ultralytics/pull/26553) by [@Y-T-G](https://github.com/Y-T-G) adds failure analysis to CI Slack alerts.
- **Shared failure alerts:** [PR #26592](https://github.com/ultralytics/ultralytics/pull/26592) by [@glenn-jocher](https://github.com/glenn-jocher) unifies CI and Docker Slack alert handling.
- **Less notification noise:** [PR #26613](https://github.com/ultralytics/ultralytics/pull/26613) by [@glenn-jocher](https://github.com/glenn-jocher) suppresses Slack notifications on workflow reruns.
- **Updated build infrastructure:** [PR #26612](https://github.com/ultralytics/ultralytics/pull/26612) by [@glenn-jocher](https://github.com/glenn-jocher) updates PyTorch to 2.14.1 and GitHub runners to 2.338.0.

## 🙌 Welcome, New Contributors!

A warm welcome to [@8rulerstar](https://github.com/8rulerstar), [@maisternia](https://github.com/maisternia), and [@shoutoutuoadi325](https://github.com/shoutoutuoadi325) for their first contributions! Thanks to the YOLO community and Ultralytics team for making this reliability-focused release possible.

## 🧪 Try v8.4.175 and Share Your Feedback

Install the release with:

```bash
pip install --upgrade ultralytics==8.4.175

```

Review the [full changelog](https://github.com/ultralytics/ultralytics/compare/v8.4.174...v8.4.175) to see everything included, then try it with your training, validation, inference, or deployment workflow.

**How does v8.4.175 work with your setup?** Share your results or report any issues below—reproduction steps and environment details are especially helpful.