Skip to content

Dataset · ibm-nasa-geospatial

Satellite Imagery for Landslide Detection

ibm-nasa-geospatial/Landslide4sense

Multispectral satellite patches with elevation data for computer vision landslide detection.

What's inside

The dataset contains 4,844 Earth surface image patches divided into training (3,799), validation (245), and test (800) splits. Each patch combines 14 channels: 12 multispectral bands from Sentinel-2, slope steepness data from ALOS PALSAR, and a digital elevation model (DEM). Originally presented in the Landslide4Sense-2022 repository.

What you can do with it

  • Training segmentation or classification models for landslide detection from satellite data
  • Comparing multispectral analysis algorithms for Earth remote sensing
Size
1K–10K rows
Downloads this month
25.5K
Likes
9

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("ibm-nasa-geospatial/Landslide4sense")