fix(deps): update dependency ultralytics ( 8.3.175 → 8.3.177 ) - #84

v8.3.177: - ultralytics 8.3.177 Add automated SPDX SBOM generation to releases (#​21643)

Compare Source

🌟 Summary

Automated SPDX SBOMs are now generated and attached to every release, improving security and compliance; plus enhanced TorchScript export flexibility with dynamic shapes and expanded MNN export testing. 🔐📦🧪

📊 Key Changes

• Automated SBOM generation in SPDX JSON for each release via a new GitHub Actions job; artifact is uploaded to the release. 🧾
• Workflow hardening: safer PR title handling to avoid shell parsing issues; actions/download-artifact updated to v5. 🔧
• TorchScript exports now support dynamic=True, enabling variable input sizes; docs, examples, and tests updated. 📐
• CI coverage expanded with an MNN export test matrix across tasks, precisions (INT8/FP16, mutually exclusive), and batches. ✅
• Docs improvements:
  • Added YouTube tutorials to COCO8-Grayscale and COCO8-Multispectral dataset pages for training YOLO11. 🎥
  • Fixed missing imports in the SAHI tiled inference guide. 🧩
  • Refined Comet callback docs and updated authors metadata. 📚
  • Polished Parking Management guide formatting. ✨
  • Updated docs banner to promote the YOLO Vision 2025 event. 🎉
• Version bump to 8.3.177. 🔖

🎯 Purpose & Impact

• Stronger supply chain transparency and compliance: SBOMs help with audits, vulnerability tracking, and enterprise procurement. 🔍✅
• More reliable release automation and CI stability, reducing risk of workflow failures. 🛡️
• Greater deployment flexibility: dynamic TorchScript supports varying image sizes without re-exporting, aligning with other formats. ♻️
• Higher confidence in MNN deployments through broader automated testing coverage. 🧪
• Faster onboarding and better UX with clearer docs, working examples, and video guides for YOLO11 and Ultralytics HUB workflows. 🚀

What's Changed

• Fix missing imports in sahi-tiled-inference guide by @​CoderUni in https://github.com/ultralytics/ultralytics/pull/21616
• Add MNN matrix export tests to --slow CI by @​lakshanthad in https://github.com/ultralytics/ultralytics/pull/21631
• Add missing dynamic=True for torchscript exports by @​lakshanthad in https://github.com/ultralytics/ultralytics/pull/21573
• Add https://youtu.be/yw2Fo6qjJU4 to docs by @​RizwanMunawar in https://github.com/ultralytics/ultralytics/pull/21633
• Remove extra - from parking-management.md FAQs by @​RizwanMunawar in https://github.com/ultralytics/ultralytics/pull/21640
• Bump actions/download-artifact from 4 to 5 in /.github/workflows by @​dependabot[bot] inhttps://github.com/ultralytics/ultralytics/pull/216399
• Docs: YOLO Vision 2025 Banner by @​sergiuwaxmann in https://github.com/ultralytics/ultralytics/pull/21623
• ultralytics 8.3.177 Add automated SPDX SBOM generation to releases by @​glenn-jocher in https://github.com/ultralytics/ultralytics/pull/21643

New Contributors

• @​CoderUni made their first contribution in https://github.com/ultralytics/ultralytics/pull/21616

Full Changelog: ultralytics/ultralytics@v8.3.176...v8.3.177

v8.3.176: - ultralytics 8.3.176 Comet integration fix for YOLO Detect and OBB models (#​21613)

Compare Source

🌟 Summary

Ultralytics 8.3.176 improves experiment tracking with Comet ML, streamlines YOLOE exports, and refactors tracking internals for cleaner, more flexible pipelines—plus a new CoreML video to speed up Apple deployments. 🚀📊

📊 Key Changes

• Comet ML integration (priority)
  • Automatically logs results.csv and args.yaml at the end of training for Detect and OBB (oriented bounding box) models. 📝
  • Adds helpers to log assets and tables; improves plot logging for detection/OBB metrics; clarifies type hints.
• YOLOE export and class-handling
  • Fuses YOLOE prompt embeddings during export only (not during generic fuse), improving compatibility. 🧩
  • Allows setting new classes even after fusion for YOLOE models (removed previous restriction).
• Trackers cleanup
  • BOT-SORT and BYTETracker now use a unified Results object (instead of separate arrays) to initialize and update tracks. 🔄
  • Reduces duplicate code and improves consistency; no changes to typical user-facing APIs.
• Docs enhancement
  • Added a YouTube tutorial to the CoreML guide showing how to export YOLO11 for faster Apple device inference. 🎥📱

🎯 Purpose & Impact

• Better experiment traceability
  • Key artifacts (metrics table, run config) are saved automatically to Comet ML, making runs easier to analyze, reproduce, and share. ✅
• Smoother YOLOE deployment
  • Export-time fusion reduces errors and increases format compatibility; post-fusion class setting makes model customization simpler. ⚙️
• More maintainable tracking
  • Unified result handling simplifies extensions and future features; developers benefit from cleaner, consistent internals. 🧼
• Faster Apple workflows
  • The new CoreML video helps users export YOLO11 for up to 2x faster inference on iOS and macOS, improving real-world deployment speed. ⚡

No expected breaking changes for most users. 🎉

What's Changed

• Add https://youtu.be/hfSK3Mk5P0I to docs by @​RizwanMunawar in https://github.com/ultralytics/ultralytics/pull/21607
• YOLOE: Scope prompt embedding fuse to export process by @​Laughing-q in https://github.com/ultralytics/ultralytics/pull/21608
• Use Results object to filter candidates for tracker by @​Laughing-q in https://github.com/ultralytics/ultralytics/pull/21536
• ultralytics 8.3.176 Comet integration fix for YOLO Detect and OBB models by @​yaricom in https://github.com/ultralytics/ultralytics/pull/21613

Full Changelog: ultralytics/ultralytics@v8.3.175...v8.3.176