mirpri/LPNSR

---
license: mit
task_categories:
- image-to-image
tags:
- image-super-resolution
- diffusion
- pytorch
configs:
- config_name: imagenet512
data_files: "imagenet512/**"
- config_name: RealSR
data_files: "RealSR/**"
---
# LPNSR: Prior-Enhanced Diffusion Image Super-Resolution Dataset

This repository contains the evaluation datasets and testing data associated with the paper **[LPNSR: Prior-Enhanced Diffusion Image Super-Resolution via LR-Guided Noise Prediction](https://huggingface.co/papers/2603.21045)**.

## Project Links

- **Paper:** [arXiv:2603.21045](https://arxiv.org/abs/2603.21045)
- **GitHub Repository:** [Faze-Hsw/LPNSR](https://github.com/Faze-Hsw/LPNSR)

## Dataset Description

This dataset collection is used to evaluate image super-resolution models on both synthetic and complex real-world degradations. It contains pairs of Low-Quality (LQ) and Ground-Truth (GT) high-resolution images.

The LPNSR approach utilizes an **LR-guided multi-input-aware noise predictor** instead of random Gaussian noise for partial diffusion initialization, allowing for efficient 4-step inference.

### Sub-Datasets Included:

- **`imagenet512`**: Contains 3,000 synthetic image pairs used for validation/testing.
- **`RealSR`**: Contains 100 image pairs featuring real-world degradations captured from actual camera sensors.
- **`RealSet80`**: Contains 80 real-world highly degraded images without Ground-Truth references.

## How to Use

Load the paired super-resolution datasets (`imagenet512`, `RealSR`) using the Hugging Face `datasets` library:

```python
from datasets import load_dataset

# Load the imagenet512 subset
dataset_imagenet = load_dataset("mirpri/LPNSR-dataset", name="imagenet512")

# Load the RealSR subset
dataset_realsr = load_dataset("mirpri/LPNSR-dataset", name="RealSR")

# Check the properties of the first pair
print(dataset_imagenet['train'][0])
# Keys will map to 'image' (for LQ) and 'ground_truth' (for GT).
```

## Citation

If you find this dataset or the LPNSR framework useful, please cite our paper:

```bibtex
@article{lpnsr2026,
title={LPNSR: Prior-Enhanced Diffusion Image Super-Resolution via LR-Guided Noise Prediction},
author={Huang, Shuwei and Liu, Shizhuo and Wei, Zijun},
journal={arXiv preprint arXiv:2603.21045},
year={2026},
eprint={2603.21045},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```

## Acknowledgement

This project is based on [ResShift](https://github.com/zsyOAOA/ResShift), [BasicSR](https://github.com/XPixelGroup/BasicSR), [SwinIR](https://github.com/JingyunLiang/SwinIR), and [Real-ESRGAN](https://github.com/xinntao/Real-ESRGAN).