YONGHUNI/GEEPatch
A Python toolkit designed to construct deep learning-ready datasets from Google Earth Engine
What's novel
A Python toolkit designed to construct deep learning-ready datasets from Google Earth Engine
Code Analysis
7 files read · 3 roundsDownloads satellite image patches from Google Earth Engine with pixel-perfect alignment on the Web Mercator grid, converting NPY data to normalized 8-bit PNGs for deep learning pipelines.
Strengths
Clean separation of concerns across geometry, processor, auth, and core modules with correct Web Mercator tile math, thorough defensive coding (atomic writes, band validation, dtype checks, NaN handling), and a well-designed two-stage concurrent pipeline that respects GEE API limits.
Weaknesses
MAX_TILES_LIMIT is defined but never enforced, creating a real safety gap; test coverage misses metadata export, deduplication, and nan_policy paths; extract_geometry silently drops all but the first feature of a FeatureCollection.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
7
Contributors
1
Files
58
Active weeks
2
Repository
Language
Python
Stars
2
Forks
0
License
MIT