On the Generalizability of Foundation Models for Crop Type Mapping

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chang, Yi-Chia, Stewart, Adam J., Bastani, Favyen, Wolters, Piper, Kannan, Shreya, Huber, George R., Wang, Jingtong, Banerjee, Arindam
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914492624928768
author Chang, Yi-Chia
Stewart, Adam J.
Bastani, Favyen
Wolters, Piper
Kannan, Shreya
Huber, George R.
Wang, Jingtong
Banerjee, Arindam
author_facet Chang, Yi-Chia
Stewart, Adam J.
Bastani, Favyen
Wolters, Piper
Kannan, Shreya
Huber, George R.
Wang, Jingtong
Banerjee, Arindam
contents Foundation models pre-trained using self-supervised learning have shown powerful transfer learning capabilities on various downstream tasks, including language understanding, text generation, and image recognition. The Earth observation (EO) field has produced several foundation models pre-trained directly on multispectral satellite imagery for applications like precision agriculture, wildfire and drought monitoring, and natural disaster response. However, few studies have investigated the ability of these models to generalize to new geographic locations, and potential concerns of geospatial bias -- models trained on data-rich developed nations not transferring well to data-scarce developing nations -- remain. We evaluate three popular EO foundation models, SSL4EO-S12, SatlasPretrain, and ImageNet, on five crop classification datasets across five continents. Results show that pre-trained weights designed explicitly for Sentinel-2, such as SSL4EO-S12, outperform general pre-trained weights like ImageNet. While only 100 labeled images are sufficient for achieving high overall accuracy, 900 images are required to mitigate class imbalance and improve average accuracy.
format Preprint
id arxiv_https___arxiv_org_abs_2409_09451
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle On the Generalizability of Foundation Models for Crop Type Mapping
Chang, Yi-Chia
Stewart, Adam J.
Bastani, Favyen
Wolters, Piper
Kannan, Shreya
Huber, George R.
Wang, Jingtong
Banerjee, Arindam
Computer Vision and Pattern Recognition
Machine Learning
Foundation models pre-trained using self-supervised learning have shown powerful transfer learning capabilities on various downstream tasks, including language understanding, text generation, and image recognition. The Earth observation (EO) field has produced several foundation models pre-trained directly on multispectral satellite imagery for applications like precision agriculture, wildfire and drought monitoring, and natural disaster response. However, few studies have investigated the ability of these models to generalize to new geographic locations, and potential concerns of geospatial bias -- models trained on data-rich developed nations not transferring well to data-scarce developing nations -- remain. We evaluate three popular EO foundation models, SSL4EO-S12, SatlasPretrain, and ImageNet, on five crop classification datasets across five continents. Results show that pre-trained weights designed explicitly for Sentinel-2, such as SSL4EO-S12, outperform general pre-trained weights like ImageNet. While only 100 labeled images are sufficient for achieving high overall accuracy, 900 images are required to mitigate class imbalance and improve average accuracy.
title On the Generalizability of Foundation Models for Crop Type Mapping
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2409.09451