# PolypGen dataset: Addressing generalisability in polyp detection and segmentation ### Part of this dataset was used in 3rd International Endoscopy Computer Vision Challenge and Workshop (EndoCV2021) [Link to EndoCV2021 webpage](https://endocv2021.grand-challenge.org/EndoCV2021/) ### **About EndoCV** Endoscopic computer vision is a challenge initiative led by [Dr Sharib Ali](https://sharibox.github.io/) designed to collaborate and curate multicenter datasets and promote building of generalisable models and assess deep learning methods. The provided dataset is an extended version of EndoCV2021 challenge. Please cite both dataset paper and challenge method paper once published (currently under preparation). ### **Data useage rules applies** [1] The data is released under the licence: **Creative Common CC BY 4.0**, which means that it will be publicly available. However, you must give appropriate credit. [2] The licensing of new creations must use the exact same terms as in the current version of the data set. [3] Should you wish to use or refer to this data set, you must cite [this paper](https://arxiv.org/pdf/2106.04463.pdf) ### **About data** - Both sequence and single frame data are provided with - Bounding box annotations in VOC format - All image and mask files are in '.jpg' format - Binary masks are provided for all single frames - Only images are provided for negative samples in the sequence data - Centerwise split is provided to encourage the development of generalisable methods. It solely depends on users how to exploit these for better generalisability tests - Folder structure with corresponding numbers are provided below **MainFolder:** *PolypGen2021_MultiCenterData* ├── codes ├── data_C1 │   ├── bbox_C1 │   ├── bbox_image_C1 │   ├── images_C1 │   └── masks_C1 ├── data_C2 │   ├── bbox_C2 │   ├── bbox_image_C2 │   ├── images_C2 │   └── masks_C2 ├── data_C3 │   ├── bbox_C3 │   ├── bbox_image_C3 │   ├── images_C3 │   └── masks_C3 ├── data_C4 │   ├── bbox_C4 │   ├── bbox_image_C4 │   ├── images_C4 │   └── masks_C4 ├── data_C5 │   ├── bbox_C5 │   ├── bbox_image_C5 │   ├── images_C5 │   └── masks_C5 ├── data_C6 │   ├── bbox_C6 │   ├── bbox_images_C6 │   ├── images_C6 │   └── masks_C6 ├── imagesAll_positive ├── dataDetails_PolypGen_SingleFrames └── sequenceData ├── negativeOnly │   ├── seq1_neg │   ├── seq2_neg │   ├── seq3_neg │   ├── seq4_neg │   ├── seq5_neg │   ├── seq6_neg │   ├── seq7_neg │   ├── seq8_neg │   └── seq9_neg │   ├── seq10_neg │   ├── seq11_neg │   ├── seq12_neg │   ├── seq13_neg │   ├── seq14_neg │   ├── seq15_neg │   ├── seq16_neg │   ├── seq17_neg │   ├── seq18_neg │   ├── seq19_neg │   ├── seq20_neg │   ├── seq21_neg │   ├── seq22_neg │   ├── seq23_neg └── positive ├── seq1 │   ├── bbox_image_seq1 │   ├── bbox_seq1 │   ├── images_seq1 │   └── masks_seq1 ├── seq2 │   ├── bbox_image_seq2 │   ├── bbox_seq2 │   ├── images_seq2 │   └── masks_seq2 ├── seq3 │   ├── bbox_image_seq3 │   ├── bbox_seq3 │   ├── images_seq3 │   └── masks_seq3 ├── seq4 │   ├── bbox_image_seq4 │   ├── bbox_seq4 │   ├── images_seq4 │   └── masks_seq4 ├── seq5 │   ├── bbox_image_seq5 │   ├── bbox_seq5 │   ├── images_seq5 │   └── masks_seq5 ├── seq6 │   ├── bbox_image_seq6 │   ├── bbox_seq6 │   ├── images_seq6 │   └── masks_seq6 ├── seq7 │   ├── bbox_image_seq7 │   ├── bbox_seq7 │   ├── images_seq7 │   └── masks_seq7 ├── seq8 │   ├── bbox_image_seq8 │   ├── bbox_seq8 │   ├── images_seq8 │   └── masks_seq8 └── seq9 ├── bbox_image_seq9 ├── bbox_seq9 ├── images_seq9 └── masks_seq9 ├── seq10 │   ├── bbox_image_seq10 │   ├── bbox_seq10 │   ├── images_seq10 │   └── masks_seq10 ├── seq11 │   ├── bbox_image_seq11 │   ├── bbox_seq11 │   ├── images_seq11 │   └── masks_seq11 ├── seq12 │   ├── bbox_image_seq12 │   ├── bbox_seq12 │   ├── images_seq12 │   └── masks_seq12 ├── seq13 │   ├── bbox_image_seq13 │   ├── bbox_seq13 │   ├── images_seq13 │   └── masks_seq13 ├── seq14 │   ├── bbox_image_seq14 │   ├── bbox_seq14 │   ├── images_seq14 │   └── masks_seq14 ├── seq15 │   ├── bbox_image_seq15 │   ├── bbox_seq15 │   ├── images_seq15 │   └── masks_seq15 ├── seq16 │   ├── bbox_image_seq16 │   ├── bbox_seq16 │   ├── images_seq16 │   └── masks_seq16 ├── seq17 │   ├── bbox_image_seq17 │   ├── bbox_seq17 │   ├── images_seq17 │   └── masks_seq17 ├── seq18 │   ├── bbox_image_seq18 │   ├── bbox_seq18 │   ├── images_seq18 │   └── masks_seq18 ├── seq19 │   ├── bbox_image_seq19 │   ├── bbox_seq19 │   ├── images_seq19 │   └── masks_seq19 ├── seq20 │   ├── bbox_image_seq20 │   ├── bbox_seq20 │   ├── images_seq20 │   └── masks_seq20 ├── seq21 │   ├── bbox_image_seq21 │   ├── bbox_seq21 │   ├── images_seq21 │   └── masks_seq21 ├── seq22 │   ├── bbox_image_seq22 │   ├── bbox_seq22 │   ├── images_seq22 │   └── masks_seq22 ├── seq23 │   ├── bbox_image_seq23 │   ├── bbox_seq23 │   ├── images_seq23 │   └── masks_seq23 174 directories **Available softwares** [1] [Data preparation and evaluation metrics](https://github.com/sharibox/EndoCV2021-polyp_det_seg_gen.git) [2] [Algorithm benchmark](https://github.com/sharibox/PolypGen-Benchmark.git) **Lead contact** [Dr. Sharib Ali](ali.sharib2002@gmail.com) **Disclaimer:** The data and information provided is for research purpose only. You should not rely upon the data provided for direct usage in software. The accuracy, reliability and completeness of the annotations may be subjective to the annotators.