README.md

---
license: cc-by-4.0
---

# 🚨 New Dataset Version Released!  

## We are excited to announce the release of **Version [2.0]** of our dataset!  
## This update includes:  
- **[More data]**.
- **[Harmonization model retrained with more data]**.
- **[Temporal support]**.
- **[Check the data without downloading (Cloud-optimized properties)]**.

# 📥 Go to: https://huggingface.co/datasets/tacofoundation/SEN2NAIPv2 and follow the instructions in colab


<center>
  <img src="demo/logo.png" width=95%>
</center>


# SEN2NAIP

The increasing demand for high spatial resolution in remote sensing imagery has led to the necessity of super-resolution (SR) algorithms that 
convert low-resolution (LR) images into high-resolution (HR) ones. To address this need, we introduce SEN2NAIP, a large remote sensing dataset
designed to support conventional and reference-based SR model training. SEN2NAIP is structured into two components to provide a broad spectrum
of research and application needs. The first component comprises a cross-sensor dataset of 2,851 pairs of LR images from Sentinel-2 L2A and HR
images from the National Agriculture Imagery Program (NAIP). Leveraging this dataset, we developed a degradation model capable of converting NAIP
images to match the characteristics of Sentinel-2 imagery (S2like). Subsequently, this degradation model was utilized to create the second component,
a synthetic dataset comprising 17,657 NAIP and S2like image pairs. With the SEN2NAIP dataset, we aim to provide a valuable resource that facilitates
the exploration of new techniques for enhancing the spatial resolution of Sentinel-2 satellite imagery.

# DOWNLOAD DATASET

```
from huggingface_hub import hf_hub_download

# Donwload cross-sensor dataset
hf_hub_download(
    repo_id="isp-uv-es/SEN2NAIP",
    repo_type="dataset",
    filename="cross-sensor/cross-sensor.zip"
)

# Donwload synthetic dataset
for i in range(1, 19):
    hf_hub_download(
        repo_id="isp-uv-es/SEN2NAIP",
        repo_type="dataset",
        filename="synthetic/synthetic_%02d.zip" % i
    )
```



# REPRODUCIBLE EXAMPLES

## Load cross-sensor dataset

```{python}
import rioxarray
import torch

DEMO_PATH = "https://huggingface.co/datasets/isp-uv-es/SEN2NAIP/resolve/main/demo/"

cross_sensor_path = DEMO_PATH + "cross-sensor/ROI_0000/"
hr_data = rioxarray.open_rasterio(cross_sensor_path + "hr.tif")
lr_data = rioxarray.open_rasterio(cross_sensor_path + "lr.tif")
hr_torch = torch.from_numpy(hr_data.to_numpy()) / 255
lr_torch = torch.from_numpy(lr_data.to_numpy()) / 10000
```


## Load Synthetic dataset

Available methods: **vae_histogram_matching**, **vae_histogram_matching**, **gamma_multivariate_normal_90**, **gamma_multivariate_normal_75**, **gamma_multivariate_normal_50**,
**gamma_multivariate_normal_25**, **gamma_multivariate_normal_10**.



```{python}
import opensr_degradation
import rioxarray
import datasets
import requests
import tempfile
import torch
import json


def load_metadata(metadata_path: str) -> dict:
    tmpfile = tempfile.NamedTemporaryFile(suffix=".json")
    with requests.get(metadata_path) as response:
        with open(tmpfile.name, "wb") as file:
            file.write(response.content)
        metadata_json = json.load(open(tmpfile.name, "r"))
    return metadata_json

DEMO_PATH = "https://huggingface.co/datasets/isp-uv-es/SEN2NAIP/resolve/main/demo/"

# Synthetic LR and HR data ------------------------------
synthetic_path = DEMO_PATH + "synthetic/ROI_0001/"

hr_early_data = rioxarray.open_rasterio(synthetic_path + "early/01__m_4506807_nw_19_1_20110818.tif")
hr_early_torch = torch.from_numpy(hr_early_data.to_numpy()) / 255
hr_early_metadata = load_metadata(synthetic_path + "late/metadata.json")
lr_hat, hr_hat = opensr_degradation.main.get_s2like(
    image=hr_early_torch,
    table=hr_early_metadata["sim_histograms"],
    model="gamma_multivariate_normal_50"
)


import matplotlib.pyplot as plt
fig, ax = plt.subplots(1, 3, figsize=(10, 5))
ax[0].imshow(hr_early_torch[[3, 1, 2]].permute(1, 2, 0))
ax[0].set_title("NAIP")
ax[1].imshow(hr_hat[[3, 1, 2]].permute(1, 2, 0)*3)
ax[1].set_title("NAIPhat")
ax[2].imshow(lr_hat[[3, 1, 2]].permute(1, 2, 0)*3)
ax[2].set_title("S2like")
plt.show()
```

<center>
  <img src="https://github.com/ESAOpenSR/opensr-degradation/assets/16768318/c88fa16e-bbe7-4072-b518-5ab3b7278893" width=100%>
</center>

# CITATION


```
@article{aybar2025sen2naipv2,
  author    = {Aybar, Cesar and Montero, David and Contreras, Julio and Donike, Simon and Kalaitzis, Freddie and Gómez-Chova, Luis},
  title     = {SEN2NAIP: A large-scale dataset for Sentinel-2 Image Super-Resolution},
  journal   = {Scientific Data},
  year      = {2024},
  volume    = {11},
  number    = {1},
  pages     = {1389},
  doi       = {10.1038/s41597-024-04214-y},
  url       = {https://doi.org/10.1038/s41597-024-04214-y},
  abstract  = {The increasing demand for high spatial resolution in remote sensing has underscored the need for super-resolution (SR) algorithms that can upscale low-resolution (LR) images to high-resolution (HR) ones. To address this, we present SEN2NAIP, a novel and extensive dataset explicitly developed to support SR model training. SEN2NAIP comprises two main components. The first is a set of 2,851 LR-HR image pairs, each covering 1.46 square kilometers. These pairs are produced using LR images from Sentinel-2 (S2) and corresponding HR images from the National Agriculture Imagery Program (NAIP). Using this cross-sensor dataset, we developed a degradation model capable of converting NAIP images to match the characteristics of S2 imagery ($S_{2-like}$). This led to the creation of a second subset, consisting of 35,314 NAIP images and their corresponding $S_{2-like}$ counterparts, generated using the degradation model. With the SEN2NAIP dataset, we aim to provide a valuable resource that facilitates the exploration of new techniques for enhancing the spatial resolution of Sentinel-2 imagery.},
  issn      = {2052-4463}
}
```