#127127th of 242 in ML frameworks and libraries

torchgeo

PyTorch library of geospatial datasets, samplers and pretrained models

Stars
4.2k
License
MIT
Last commit
Nov 2026
Last release
Sep 2026
Language
Python
Stars, last 29 weeks+270
25 Apr 2026: 3,973 stars2 May 2026: 3,975 stars9 May 2026: 3,989 stars16 May 2026: 4,006 stars23 May 2026: 4,019 stars30 May 2026: 4,029 stars6 Jun 2026: 4,037 stars13 Jun 2026: 4,048 stars20 Jun 2026: 4,054 stars27 Jun 2026: 4,066 stars4 Jul 2026: 4,074 stars11 Jul 2026: 4,089 stars18 Jul 2026: 4,096 stars25 Jul 2026: 4,100 stars1 Aug 2026: 4,111 stars8 Aug 2026: 4,123 stars15 Aug 2026: 4,131 stars22 Aug 2026: 4,143 stars29 Aug 2026: 4,150 stars5 Sept 2026: 4,161 stars12 Sept 2026: 4,167 stars19 Sept 2026: 4,174 stars26 Sept 2026: 4,179 stars3 Oct 2026: 4,182 stars10 Oct 2026: 4,187 stars17 Oct 2026: 4,197 stars24 Oct 2026: 4,204 stars31 Oct 2026: 4,214 stars7 Nov 2026: 4,232 stars14 Nov 2026: 4,243 stars

3,973 on 25 Apr 20264,243 on 14 Nov 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

TorchGeo is a PyTorch domain library, modeled on torchvision, for training models on remote sensing data. It provides geospatial datasets that match CRS and resolution, samplers that draw patches by coordinates, benchmark datasets with download and checksum, and pretrained weights for multispectral sensors. Lightning datamodules, tasks and a CLI cover standard training runs.

Who it is for: ML engineers and remote sensing researchers training models on satellite imagery

Strengths

  • Union (|) and intersection (&) operators combine datasets with differing CRS and resolution
  • Pretrained weights for multispectral sensors, using torchvision's multi-weight API
  • Lightning datamodules with fixed splits and a YAML-driven `torchgeo` CLI
  • Benchmark datasets download and checksum like torchvision, returning dicts of tensors

Weaknesses

  • Geospatial imagery such as Landsat must be downloaded by the user; only some datasets auto-download
  • Tied to PyTorch and Lightning; no support for other tensor frameworks
  • Library only, with no UI or hosted service
  • Built-in tasks cover classification, regression and segmentation; other tasks need custom code

What it needs

  • GPU optional
  • Needs PyTorch, Lightning, kornia, timm
  • Models: PyTorch, U-Net, ResNet, Mask R-CNN

What changed

  1. v0.10.0 ↗ (opens in a new tab)Security fix

    “LLMs have made keeping up with security... a fun challenge. This release includes a number of important steps to harden security across TorchGeo:”

From the release notes. No flag means the note does not say so, not that the upgrade is safe.

Also in ML frameworks and libraries

See all 242
Also in ML frameworks and libraries
RankProjectScore
1transformersModel-definition library for pretrained text, vision, audio and multimodal models169.3k starsApache-2.094 out of 100
2ultralyticsPython library and CLI for training and running YOLO vision models63.5k starsAGPL-3.094 out of 100
3pytorch-image-modelsPyTorch library of image models, pretrained weights and training scripts37.3k starsApache-2.094 out of 100
#4pytorchPython tensor library with GPU support and tape-based autograd105.8k starscustom license93 out of 100
#5supervisionPython toolkit for loading, annotating and counting computer vision detections Live demo ↗ (opens in a new tab)51.7k starsMIT92 out of 100