Datasets:

Angelou0516
/

polyp-seg-generalization-bench

license: cc-by-4.0
task_categories:
  - image-segmentation
tags:
  - medical
  - polyp-segmentation
  - colonoscopy
  - gastrointestinal
  - generalization
  - benchmark
size_categories:
  - 1K<n<10K
pretty_name: Polyp Segmentation Generalization Benchmark

Polyp Segmentation Generalization Benchmark

Dataset Description

This benchmark dataset is designed for evaluating cross-domain generalization in polyp segmentation. It contains training data from mixed Kvasir-SEG and CVC-ClinicDB sources, with multiple test sets from different colonoscopy databases for comprehensive evaluation.

Task: Binary segmentation (polyp vs. background)
Modality: Colonoscopy/Endoscopy
Format: PNG images with binary masks

Dataset Structure

polyp-seg-generalization-bench/
├── train/               # Mixed Kvasir + ClinicDB training data
│   ├── images/
│   └── masks/
├── test_cvc300/         # CVC-300 test set
│   ├── images/
│   └── masks/
├── test_cvc_clinicdb/   # CVC-ClinicDB test set
│   ├── images/
│   └── masks/
├── test_cvc_colondb/    # CVC-ColonDB test set
│   ├── images/
│   └── masks/
├── test_etis/           # ETIS-LaribPolypDB test set
│   ├── images/
│   └── masks/
└── test_kvasir/         # Kvasir test set
    ├── images/
    └── masks/

Splits

Split	Description	Source
train	Mixed training data	Kvasir-SEG + CVC-ClinicDB
test_cvc300	CVC-300 test set	CVC-300
test_cvc_clinicdb	CVC-ClinicDB test set	CVC-ClinicDB
test_cvc_colondb	CVC-ColonDB test set	CVC-ColonDB
test_etis	ETIS test set	ETIS-LaribPolypDB
test_kvasir	Kvasir test set	Kvasir-SEG

Citation

If you use this dataset, please cite the original datasets:

@inproceedings{jha2020kvasir,
  title={Kvasir-SEG: A Segmented Polyp Dataset},
  author={Jha, Debesh and others},
  booktitle={MMM},
  year={2020}
}

@article{bernal2015wm,
  title={WM-DOVA maps for accurate polyp highlighting in colonoscopy},
  author={Bernal, Jorge and others},
  journal={Computerized Medical Imaging and Graphics},
  year={2015}
}

@article{tajbakhsh2015automated,
  title={Automated polyp detection in colonoscopy videos using shape and context information},
  author={Tajbakhsh, Nima and others},
  journal={IEEE TMI},
  year={2015}
}

@article{silva2014toward,
  title={Toward embedded detection of polyps in WCE images for early diagnosis of colorectal cancer},
  author={Silva, Juan and others},
  journal={IJCARS},
  year={2014}
}

Generalization Benchmark Reference:

@inproceedings{fan2020pranet,
  title={PraNet: Parallel Reverse Attention Network for Polyp Segmentation},
  author={Fan, Deng-Ping and others},
  booktitle={MICCAI},
  year={2020}
}

```bibtex
@article{chang2024esfpnet,
  title={ESFPNet: Efficient Stage-Wise Feature Pyramid on Mix Transformer for Deep Learning-Based Cancer Analysis in Endoscopic Video},
  author={Chang, Qi and Ahmad, Danish and Toth, Jennifer and Bascom, Rebecca and Higgins, William E},
  journal={Journal of Imaging},
  volume={10},
  number={8},
  pages={191},
  year={2024},
  publisher={MDPI}
}


## Usage

```python
from datasets import load_dataset

# Load dataset
dataset = load_dataset("Angelou0516/polyp-seg-generalization-bench")

# Access splits
train_data = dataset['train']
test_cvc300 = dataset['test_cvc300']
test_etis = dataset['test_etis']

# Access a sample
sample = train_data[0]
image = sample['file_name']  # Image
mask = sample['mask_file_name']  # Segmentation mask

License

Please refer to the original dataset licenses and citation requirements.