EndoBoost: a plug-and-play module for false positive suppression during computer-aided polyp detection in real-world colonoscopy (with dataset)

This is the official repo for FPPD-13 dataset and EndoBoost. It provides FPPD-13 dataset download links, preprocessing scripts and implementation of EndoBoost.

[arXiv] [Project Page]

FPPD-13 Dataset

Download FPPD-13 dataset: [GoogleDrive] [BaiduCloud] (code: fppd)

Dataset Organization

The FPPD-13 dataset is organized in the following way:

├── 0
│   ├── data
│   └── patch
├── 1
│   ├── data
│   └── patch
...
├── 13
│   ├── data
│   └── patch
├── class_id_name.txt
├── fold1
│   ├── test.csv
│   ├── train.csv
│   ...
│   └── val.csv
├── fold2
├── fold3
├── fold4
└── fold5

• class_id_name.txt provides IDs and names of true positives and all 13 classes of false positives.
• 0 to 13 are colonoscopy images in two formats:
  • patch: In this folder, image patch cropped with the prediction bounding box are provided. Recommended for reproducing experiment results.
  • data: Original images (frames of colonoscopy video) and bounding boxes in .json are provided.
• fold1 to fold5 contains dataset split files in .csv format for all experiments.

EndoBoost

Implementation of EndoBoost will be publicly available soon!

Citation

If you use the FPPD-13 dataset or the EndoBoost code please cite:

@article{wang2022endoboost,
  title={EndoBoost: a plug-and-play module for false positive suppression during computer-aided polyp detection in real-world colonoscopy (with dataset)},
  author={Wang, Haoran and Zhu, Yan and Qin, Wenzheng and Zhang, Yizhe and Zhou, Pinghong and Li, Quanlin and Wang, Shuo and Song, Zhijian},
  journal={arXiv preprint arXiv:2212.12204},
  year={2022}
}