BioSR Dataset (biosr_dataset)

Overview

This DeepTrackAI repository provides a preprocessed part of the BioSR dataset, available from figshare (DOI: 10.6084/m9.figshare.13264793.v9) and originally published by Chang Qiao et al., Nature Methods, 2021.

The original dataset consists of paired low-resolution (LR) and high-resolution (HR) fluorescence microscopy images for training and benchmarking super-resolution reconstruction methods, covering four biology structures (Clatrin Coated Pits, Endoplasmatic Reticulum, Microtubules, F-actin), nine signal levels (15-600 average photon count), and two upscaling-factors (linear SIM and non-linear SIM).

This repo only includes the Microtubules folder and the images in this repository have been cropped into 128 × 128 pixel patches, saved as 32-bit grayscale TIFF files, and organized into training/validate/test splits to be directly usable in deep learning workflows, while preserving the original content and licensing terms.

Summary

• Number of images:
  • Training: 41,040 pairs
  • Validation: 2,160 pairs
  • Test: 150 HR images × 9 LR signal levels (1,350 LR images total). Each subfolder corresponds to a different signal-to-noise level, reflecting increasing average photon counts (15–600 photons).
• Image size: 128 × 128 pixels
• Image format: 32-bit grayscale TIFF

---

Original Source

• Title: BioSR: a biological image dataset for super-resolution microscopy
• Authors: Chang Qiao and Di Li
• Source: figshare (DOI: 10.6084/m9.figshare.13264793.v9)
• Reference article: Qiao C, Li Y, Qu J, et al. Nature Methods 18: 194–202 (2021). DOI: 10.1038/s41592-020-01048-5
• License: Creative Commons Attribution 4.0 International (CC BY 4.0)

If you use this dataset in your research, please follow the licensing requirements and properly attribute the original authors.

---

Dataset Structure

/biosr_dataset
└── BioSR/
    └── Microtubules/
        ├── training_wf/      # Low-resolution training images
        │   ├── 00000001.tif
        │   ├── 00000002.tif
        │   └── ...
        ├── training_gt/      # High-resolution training images
        │   ├── 00000001.tif
        │   ├── 00000002.tif
        │   └── ...
        ├── validate_wf/      # Low-resolution validation images
        │   ├── 00000001.tif
        │   ├── 00000002.tif
        │   └── ...
        ├── validate_gt/      # High-resolution validation images
        │   ├── 00000001.tif
        │   ├── 00000002.tif
        │   └── ...
        ├── test_wf/          # Low-resolution test images, 9 signal levels
        │   ├── level_01/     # Lowest photon count (~15), lowest SNR
        │   │   ├── 001.tif
        │   │   ├── 002.tif
        │   │   └── ...
        │   ├── level_02/
        │   │   ├── 001.tif
        │   │   ├── 002.tif
        │   │   └── ...
        │   ├── ...
        │   └── level_09/     # Highest photon count (~600), highest SNR
        └── test_gt/          # High-resolution test images
            ├── 001.tif
            ├── 002.tif
            └── ...

Each filename is a sequential numerical identifier. In the test set, subfolders correspond to different signal levels (photon counts), with level_01 being the noisiest and level_09 the cleanest.

---

How to Access the Data

Clone the Repository

git clone https://github.com/DeepTrackAI/biosr_dataset
cd biosr_dataset

---

Attribution

If you use this dataset, please cite both the BIOSR dataset and the reference article.

Cite the dataset:

Qiao, Chang; Li, Di. BioSR: a biological image dataset for super-resolution microscopy. figshare. Dataset (2020). DOI: 10.6084/m9.figshare.13264793.v9

@misc{qiao2020biosr,
author = "Chang Qiao and Di Li",
title = "{BioSR: a biological image dataset for super-resolution microscopy}",
year = "2020",
month = "11",
url = "https://figshare.com/articles/dataset/BioSR/13264793",
doi = "10.6084/m9.figshare.13264793.v9"
}

Cite the reference article:

Qiao C et al. Evaluation and development of deep neural networks for image super-resolution in optical microscopy. Nature Methods, 18: 194–202 (2021). DOI: 10.1038/s41592-020-01048-5

@article{qiao2021eval,
  title={Evaluation and development of deep neural networks for image super-resolution in optical microscopy},
  author={Qiao, Chang and Li, Di and Guo, Yuting and Liu, Chong and Jiang, Tao and Dai, Qionghai and Li, Dong},
  journal={Nature Methods},
  volume={18},
  pages={194--202},
  year={2021},
  publisher={Nature Publishing Group},
  doi={10.1038/s41592-020-01048-5}
}

---

License

This replication dataset is shared under the Creative Commons Attribution 4.0 International (CC BY 4.0) License, consistent with the original licensing terms.