TerraKit: Standardising AI-Ready Geospatial Data...
summary
An open‑source Python library called TerraKit automates retrieval, processing, alignment and spatial train/val splitting of EO raster/vector labels into AI‑ready datasets from providers like EarthData, Sentinel Hub, Copernicus and Planetary Computer via a single config, producing provenance and exports such as TACO datasets for direct use with TorchGeo/TerraTorch. A wildfire case builds a multimodal dataset (optical + radar) in “five lines of code,” turning a months‑long pipeline into minutes and handling data at the 0.5 exabyte scale.
AI-written summary · about
chapters (9)
- 0:00TerraKit intro and geospatial AI
- 1:28Geospatial foundation models
- 3:57Massive dataset challenges
- 4:18What is TerraKit
- 5:22AI-ready datasets with TerraKit
- 7:07Fine-tuning geospatial models
- 8:52TerraKit use cases and future
- 9:34Closing and call to action
- 10:12Q&A
related talks
31mFrom Radar Echoes To Real-Time Recognition: Deep Learning on Another Planet -An Open Source Workflow
45mFrom Radar Echoes To Real-Time Recognition: Deep Learning on Another Planet an Open Source Workflow
48mTaking the planet’s pulse: AI to empower adaptation to global change
16mThe Open Data Cube and working towards a Digital Earth
18mGeospatial AI: Where Can AI Help to Tackle Today’s Real-World Challenges?
34mGeoscience Big Data Computing with Python
39mSatellite Imagery to Machine-Readable Intelligence
15mChallenges in Practice: Building a Usable Library for Planetary-Scale Embeddings