What it does

This is a curated list—think of it as a bibliography with links—covering machine learning applied to geospatial data science. It rounds up code projects, datasets, academic papers, books, courses, and a lone company working on satellite imagery analysis, semantic segmentation, and remote sensing classification.

The interesting bit

The field is genuinely fragmented: relevant work sits in computer vision journals, GIS conferences, Kaggle competitions, and random GitHub repos. The maintainer has done the tedious work of gathering U-Net implementations for satellite segmentation, vegetation forecasting with RNNs, and hyperspectral classification papers into one place.

Key highlights

• Code projects span practical workflows: label preparation (Development Seed’s label-maker), DeepLab v3 and U-Net adaptations for satellite imagery, YOLO-inspired object detection, and raster-vision for aerial deep learning pipelines
• Datasets include well-known benchmarks: SpaceNet commercial imagery corpus, Dstl’s 1km multi-band satellite tiles, and DeepSat’s 405,000 land-cover patches from NAIP aerial photography
• Papers cover 2016–2018 vintage: hyperspectral CNNs on GPU, TensorFlow systems papers, and remote sensing scene classification with data augmentation
• Books and courses are included: standard Springer remote sensing texts, Google’s ML crash course, and Stanford’s TensorFlow for Deep Learning Research
• 702 stars suggests modest but steady interest from researchers crossing over from either ML or GIS backgrounds

Caveats

• Last substantial update appears to be 2018; many links and frameworks (TensorFlow 1.x-era code, old Kaggle competitions) may be stale
• Coverage is broad rather than deep: no ranking, no curation notes on code quality or reproducibility, and no distinction between actively maintained projects and abandoned experiments
• “Companies” section contains exactly one entry (SpaceKnow), suggesting the category is aspirational rather than comprehensive

Verdict

Worth bookmarking if you’re a grad student or researcher entering geospatial ML and need a map of the territory circa 2018. Skip it if you want maintained code, modern PyTorch implementations, or critical evaluation of which approaches actually work in production.