M0: Environment and project setup¶

Status: done · Commit: c8c701a

Goal¶

A reproducible Python project with the computer-vision stack verified, and a written plan grounded in the project references.

Reviewed the references (docs/references.md):
LessWrong, "Computational Thread Art": a greedy line choice, a penalty with ± importance weights and a darkness term D, per-colour Floyd–Steinberg dithering, and numpy speed-ups (~10 s).
IEEE 10844087, "Automated String Art Creation": mostly the fabrication machine (CNC, 3D printing, Arduino) plus a CNN. It's useful as fabrication context, not as an algorithm baseline.
Added Birsak et al. 2018 (CGF) as the academic baseline. It compares a fine-scale canvas against a coarse target, which models viewing distance.
Created a uv package project (src/ layout), with Python 3.12 pinned in .python-version.
Wrote the project plan: improvements I1–I7, architecture, milestones, evaluation protocol, risks, and the visualizer spec (§2b).

Item	Version / note
uv	0.11.2
Python	3.12.13 (pinned; widest wheel support for numba and OpenCV)
opencv-contrib-python	5.0.0. Do not also install `opencv-python`; they conflict
numpy / scipy / scikit-image	2.5.3 / 1.18.1 / 0.26.0
numba	0.68.0 (JIT verified)
dev	pytest 9.1, ruff 0.16

Verified OpenCV modules: FaceDetectorYN (YuNet), face.createFacemarkLBF, saliency, ximgproc, createCLAHE.

ruff.exe is blocked by Windows Smart App Control ("An Application Control policy has blocked this file"), so linting can't run on the dev machine. Workarounds: allow the binary, or run uv tool install ruff.

uv sync
uv run python -c "import cv2; print(cv2.__version__, hasattr(cv2, 'FaceDetectorYN'))"