ultralytics 8.3.225

v8.3.225 - `ultralytics 8.3.225` New KITTI dataset (#22539)

on Python PyPI

🌟 Summary

Ultralytics 8.3.225 adds first-class KITTI dataset support for YOLO11, improves model checkpoint loading reliability, cleans up docs/tests, and makes Jetson builds more reproducible. 🚗🧠⚙️

📊 Key Changes

• New: KITTI dataset support and docs 🎉

  • Added kitti.yaml with classes and download link.
  • New documentation page with Python/CLI examples and MkDocs navigation entries.
  • See the KITTI dataset documentation and the kitti.yaml config.

• Safer checkpoint loading for SAM/SAM2 🛡️

  • Unified torch_load helper used for SAM and SAM2 checkpoint loading for better cross-version PyTorch compatibility.

• SAM-2 interactive predictor now uses 0-based object IDs (breaking change) 🔁

  • Standardized to 0-based IDs and updated docs/examples.

• Jetson JetPack 4 Docker build reliability 🔧

  • Switches ONNX Runtime/TensorRT wheels to Ultralytics-hosted assets for stable, reproducible builds.

• Docs polish and CI stability 📚✅

  • Fixed ExecuTorch docs code block formatting for clean rendering and copy-paste.
  • Removed flaky ExecuTorch export CLI test to stabilize CI.
  • Added comprehensive type hints and docstring improvements across dataloaders and utilities.

🎯 Purpose & Impact

• Faster start on autonomous driving research 🚦

  • Out-of-the-box KITTI support lets users train/evaluate YOLO11 on a key driving dataset with minimal setup.

• More reliable model loading across environments 🔄

  • Centralizing checkpoint loading reduces pickle/compatibility pitfalls across PyTorch versions for SAM/SAM2 users.

• Clearer APIs and fewer indexing mistakes 🧭

  • 0-based IDs in SAM-2 align with Python norms, reducing off-by-one errors. Note: update scripts from IDs starting at 1 to start at 0.

• Reproducible hardware builds 🧪

  • Jetson images are less likely to break due to dead links or external host changes.

• Better developer and user experience ✨

  • Cleaner docs and stronger typing improve readability, IDE support, and onboarding.

Quick start on KITTI with YOLO11:

yolo detect train data=kitti.yaml model=yolo11n.pt epochs=100 imgsz=640

from ultralytics import YOLO
model = YOLO("yolo11n.pt")
model.train(data="kitti.yaml", epochs=100, imgsz=640)

What's Changed

• Fix 429 forum links in Dockerfile-jetson-jetpack4 by @glenn-jocher in #22576
• Use torch_load patch for weights_only cases by @Laughing-q in #22577
• Fix indentation in integrations/executorch.md examples by @Laughing-q in #22581
• Remove redundant Executorch test by @Laughing-q in #22582
• Add Python type hints for dataloader functions by @Laughing-q in #22580
• Fix object ID indexing with SAM2DynamicInteractivePredictor by @Y-T-G in #22529
• ultralytics 8.3.225 New KITTI dataset by @RizwanMunawar in #22539

Full Changelog: v8.3.224...v8.3.225