Implement ilabel: fast pixel-based region labelling (from the MATLAB MVTB C extension)
维护者通常 1 天内回复
还没有人认领这个 Issue。
评估
- 难度
- 5/5
- 预计耗时
- 一周以上
- 新手友好度
- 25/100
- Issue 类型
- 功能
- 描述清晰度
- 需要澄清
- 活跃度
- 活跃
调研方向
Start by inspecting wip/ilabel/ilabel.c and wip/ilabel/ilabel.py-unopt, then read the existing blob/label functions and the TODO in tests/test_image_processing.py. Decide and document the Python API and implementation route, with tests comparing shared examples to the contour-based blob code. Done means ilabel is implemented, its return values are tested, and the build and Python-only support implications are addressed.
由索引模型根据 Issue 内容生成。
描述
Parked work, to be implemented later. The files are on branch feat/fast-label under wip/ilabel/ (not in main).
What it is
A pixel-based, rather than contour-based, approach to segmentation / region labelling, based on the ilabel C extension (Copyright 1995-2009 Peter Corke) from the MATLAB Machine Vision Toolbox. The MATLAB signature was [l, m, parents, color] = ilabel(im).
What exists
wip/ilabel/ilabel.c: Cython 3.0.11 output (2024-12), 592 KB. It was generated from anilabel.pyxthat is not in the repo, but Cython embeds the source as/* "ilabel.pyx":NN */comments, so the.pyxcan be reconstructed from this file.wip/ilabel/ilabel.py-unopt: unoptimised pure-Python port, produced with Copilot (2024-12).- The Python toolbox has no implementation. The only reference is a TODO in
tests/test_image_processing.py(# TODO [l,ml,p,c] = ilabel(im);).
To do
- Reconstruct
ilabel.pyxfrom the embedded comments inilabel.c(or port afresh), and decide the implementation route (Cython extension vs numba vs NumPy/SciPy). Note this package is pure Python and builds with hatchling, so a compiled extension changes the build/wheel story (and Pyodide/JupyterLite support); a pure-Python/NumPy route avoids that. - Define the Python API (
Imagemethod, return values), consistent with the existing blob/label functions. - Tests, including a comparison with the existing contour-based blob code on shared examples, and the TODO above.
Related: #45
- 主要语言
- Python
- 星标
- 223
- 派生
- 30
- 平均合并
- 1 小时 4 分钟
- 30 天内合并 PR
- 9
环境准备
- 没有 Dockerfile 或 Docker Compose 文件
- 有 Pull Request 模板
- 阅读贡献指南
从这里开始
- 先读完整个 Issue，再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库，在一个分支上完成修改。
- 提交 Pull Request，并在描述里引用这个 Issue 编号。
petercorke/machinevision-toolbox-python 的其他 Issue
-
enhancement
难度 2/5 1-3 小时 新手友好度 25/100
petercorke/machinevision-toolbox-python#125 ·
维护者通常 1 天内回复
-
tech-debt
难度 5/5 一周以上 新手友好度 25/100
petercorke/machinevision-toolbox-python#122 ·
维护者通常 1 天内回复
-
tech-debt
难度 5/5 一周以上 新手友好度 30/100
petercorke/machinevision-toolbox-python#117 ·
维护者通常 1 天内回复
-
tech-debt
难度 5/5 一周以上 新手友好度 45/100
petercorke/machinevision-toolbox-python#115 ·
维护者通常 1 天内回复
-
tech-debt
难度 4/5 3-5 天 新手友好度 62/100
petercorke/machinevision-toolbox-python#113 ·
维护者通常 1 天内回复
查看 petercorke/machinevision-toolbox-python 的全部 Issue
相似的 Issue
-
难度 2/5 1-3 小时 新手友好度 82/100
RedHatQE/mtv-api-tests#721 ·
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 84/100
维护者通常 1 天内回复
-
难度 1/5 1-3 小时 新手友好度 85/100
pytest-dev/pluggy#757 ·
维护者通常 1 天内回复
-
难度 1/5 1-3 小时 新手友好度 85/100
NousResearch/hermes-agent#134960 ·
维护者通常 1 天内回复
-
HTML backend: `<br>` leaks the internal sentinel U+E000 into list items, headings and captions可能已有人在做 @morten-lagabote 今天认领。 未关闭
难度 2/5 1-3 小时 新手友好度 67/100
docling-project/docling#4671 ·
维护者通常 1 天内回复