metadata
license: apache-2.0
language:
- en
size_categories:
- n<1K
Demo dataset for our CVPR 2025 paper "ZoomLDM: Latent Diffusion Model for multi-scale image generation". We extract patches from the Chesapeake land cover dataset.
Usage
from datasets import load_dataset
ds = load_dataset("StonyBrook-CVLab/ZoomLDM-demo-dataset-NAIP", name="3x", trust_remote_code=True, split='train')
print(np.array(ds[0]['ssl_feat']).shape)
>>> (1024, 4, 4)
Citations
@inproceedings{yellapragada2025zoomldm,
title={ZoomLDM: Latent Diffusion Model for multi-scale image generation},
author={Yellapragada, Srikar and Graikos, Alexandros and Triaridis, Kostas and Prasanna, Prateek and Gupta, Rajarsi and Saltz, Joel and Samaras, Dimitris},
booktitle={Proceedings of the Computer Vision and Pattern Recognition Conference},
pages={23453--23463},
year={2025}
}
@inproceedings{robinson2019large,
title={Large Scale High-Resolution Land Cover Mapping With Multi-Resolution Data},
author={Robinson, Caleb and Hou, Le and Malkin, Kolya and Soobitsky, Rachel and Czawlytko, Jacob and Dilkina, Bistra and Jojic, Nebojsa},
booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
pages={12726--12735},
year={2019}
}