Model Card for UNet-5depth-shuffle: venkatesh-thiru/s2l8h-UNet-5depth-shuffle

Model Description

The UNet-5depth-shuffle model aims to harmonize Landsat-8 and Sentinel-2 imagery by enhancing the spatial resolution of Landsat-8 images. The model processes Landsat-8 multispectral and pan-chromatic images to output images with Sentinel-2-like spectral and spatial characteristics.

Model Architecture

This UNet architecture includes 5 depth levels and uses a shuffling mechanism to improve image resolution and spectral alignment. The combination of depth levels and shuffling is optimized to produce high-quality images that closely match Sentinel-2 data.

Usage

from transformers import AutoModel

# Load the UNet-5depth-shuffle model
model = AutoModel.from_pretrained("venkatesh-thiru/s2l8h-UNet-5depth-shuffle", trust_remote_code=True)

# Harmonize Landsat-8 images
l8up = model(l8MS, l8pan)

Where:

l8MS - Landsat Multispectral images (L2 Reflectances) l8pan - Landsat Pan-Chromatic images (L1 Reflectances)

Potential Applications

Water quality assessment Urban planning Climate monitoring Disaster response Infrastructure oversight Agricultural surveillance

Limitations

The model shows minor limitations in areas with unique spectral characteristics or under extreme environmental conditions.

Reference

For more details, refer to the publication: 10.1016/j.isprsjprs.2024.04.026