fix(deps): update dependency ultralytics ( 8.3.177 → 8.3.178 ) - #92

v8.3.178: - ultralytics 8.3.178 new lighter Dockerfile-python images (#​21661)

Compare Source

🌟 Summary

Lighter, faster Docker images with a clear split between a minimal Python runtime and a full export toolkit, plus improved build workflows, headless deps, and small API enhancements for smoother installs and preprocessing. 🚀🐳

📊 Key Changes

• Docker images overhaul (priority)
  • New split: lightweight latest-python for inference ⚡ and full latest-python-export for model export (Edge TPU, NCNN, IMX, Paddle) 🧰
  • CI builds derivatives automatically (runner, jupyter, cpu, python-export) from base images; unified push flow 🏗️
  • Tests/benchmarks run against latest-python-export when appropriate to ensure export paths are validated ✅
  • Image descriptions/tags clarified; generalized “YOLO11” → “YOLO”; consistent handling across GPU and Jetson images 🏷️
• Smaller, more compatible images
  • Switched to opencv-python-headless and default-jre-headless in Docker builds 🧠
  • Consolidated RUN steps, added cleanup, and applied security updates; faster builds with fewer layers 🧹🔒
• Export and backend reliability
  • NCNN installs now use --no-deps to avoid pulling opencv-python; mirrors in autobackend loader 🔧
  • Docs/comments include RKNN export artifact name yolo11n_rknn_model 📘
• Preprocessing flexibility
  • LetterBox now supports padding_value and interpolation parameters for finer control over resize/pad behavior 🎨
• Supply chain improvements
  • SBOM generation scoped to an isolated “sbom-env” via uv sync; Anchore scans that env and outputs SPDX JSON 🔒📦
• Version bump
  • Ultralytics package updated to 8.3.178 🔖

🎯 Purpose & Impact

• Faster pulls and quicker startup for most users by defaulting to a slim inference image; heavy export deps moved to a dedicated image ⚡📦
• More reliable CI and clearer image taxonomy reduce confusion and improve maintenance, especially across GPU/Jetson variants 🧭
• Leaner, headless images cut download size, build times, and potential GUI-related conflicts; improved security posture ⏱️🔐
• Smoother NCNN workflows and fewer dependency conflicts, especially on minimal or ARM64 environments 🧩
• Better preprocessing control can improve model performance across diverse data types (e.g., thermal, grayscale, custom backgrounds) 🎯
• Clearer SBOMs enhance transparency and compliance for organizations using Ultralytics in production 🛡️

Enjoy faster, cleaner Docker workflows and more configurable pipelines—whether you’re prototyping with YOLO or exporting to edge backends. 🚀✨

What's Changed

• Scope publish.yml SBOM by @​glenn-jocher in https://github.com/ultralytics/ultralytics/pull/21644
• Install NCNN with '--no-deps' by @​glenn-jocher in https://github.com/ultralytics/ultralytics/pull/21648
• Use opencv-python-headless in Dockerfiles by @​glenn-jocher in https://github.com/ultralytics/ultralytics/pull/21647
• Support padding_value and interpolation in LetterBox for better compatibility by @​Laughing-q in https://github.com/ultralytics/ultralytics/pull/21652
• Combine Dockerfile RUN commands by @​glenn-jocher in https://github.com/ultralytics/ultralytics/pull/21649
• Refactor derivative Dockerfile images by @​glenn-jocher in https://github.com/ultralytics/ultralytics/pull/21656
• ultralytics 8.3.178 new lighter Dockerfile-python images by @​glenn-jocher in https://github.com/ultralytics/ultralytics/pull/21661

Full Changelog: ultralytics/ultralytics@v8.3.177...v8.3.178