Related Experiment Video
Updated: Jun 4, 2025
14:02
Super-Resolution Imaging to Study Co-Localization of Proteins and Synaptic Markers in Primary Neurons
Published on: October 31, 2020
5.7K
SEN2NAIP: A large-scale dataset for Sentinel-2 Image Super-Resolution
Cesar Aybar1, David Montero2, Julio Contreras3
1Image Processing Laboratory, University of Valencia, Valencia, Spain. cesar.aybar@uv.es.
Scientific Data
|December 18, 2024
Summary
We introduce SEN2NAIP, a new dataset for training super-resolution (SR) models. This resource aids in enhancing the spatial resolution of Sentinel-2 (S2) satellite imagery for remote sensing applications.
Area of Science:
- Remote Sensing
- Computer Vision
Background:
- High spatial resolution is crucial for remote sensing.
- Super-resolution (SR) algorithms are needed to upscale low-resolution (LR) images to high-resolution (HR).
Purpose of the Study:
- To introduce SEN2NAIP, a novel and extensive dataset for training SR models.
- To provide a valuable resource for enhancing Sentinel-2 (S2) imagery spatial resolution.
Main Methods:
- Developed a dataset of 2,851 LR-HR image pairs from Sentinel-2 (S2) and National Agriculture Imagery Program (NAIP).
- Created a degradation model to convert NAIP images to S2 characteristics.
- Generated a second subset of 35,314 NAIP images and corresponding S2-like counterparts.
Main Results:
- The SEN2NAIP dataset consists of two subsets, totaling 38,165 image pairs.
- A cross-sensor degradation model was successfully developed and applied.
- The dataset facilitates SR model training for Sentinel-2 imagery.
Conclusions:
- SEN2NAIP is a comprehensive resource for advancing SR techniques in remote sensing.
- The dataset supports research in improving the spatial resolution of Sentinel-2 data.
