The Kvasir-SEG dataset is an open-access dataset of gastrointestinal polyp images and corresponding segmentation masks, manually annotated by a medical doctor and verified by an experienced gastroenterologist. It also includes bounding boxes of the polyp regions. The dataset is designed to assist researchers in reproducing results and comparing methods in polyp segmentation and automatic analysis of colonoscopy images. The dataset extends the Kvasir dataset by adding segmentation masks and bounding boxes, enabling multimedia and computer vision researchers to contribute to the field. The dataset includes polyp images and their corresponding segmentation masks, with the ROIs (regions of interest) representing polyp tissue. The dataset is publicly available and open access. The paper presents a baseline model for evaluation and demonstrates the use of the dataset with traditional segmentation approaches and modern deep-learning based Convolutional Neural Network (CNN) approaches. The results show that the ResUNet model outperforms the FCM clustering algorithm in segmenting polyp pixels. The dataset is useful for both training and validation, and can assist in developing state-of-the-art solutions for colonoscopy images. The paper also discusses related work, including other available polyp datasets, and highlights the need for open-access datasets for comparable evaluations. The Kvasir-SEG dataset is released as open-source to the multimedia and medical research communities to help evaluate and compare existing and future computer vision methods.The Kvasir-SEG dataset is an open-access dataset of gastrointestinal polyp images and corresponding segmentation masks, manually annotated by a medical doctor and verified by an experienced gastroenterologist. It also includes bounding boxes of the polyp regions. The dataset is designed to assist researchers in reproducing results and comparing methods in polyp segmentation and automatic analysis of colonoscopy images. The dataset extends the Kvasir dataset by adding segmentation masks and bounding boxes, enabling multimedia and computer vision researchers to contribute to the field. The dataset includes polyp images and their corresponding segmentation masks, with the ROIs (regions of interest) representing polyp tissue. The dataset is publicly available and open access. The paper presents a baseline model for evaluation and demonstrates the use of the dataset with traditional segmentation approaches and modern deep-learning based Convolutional Neural Network (CNN) approaches. The results show that the ResUNet model outperforms the FCM clustering algorithm in segmenting polyp pixels. The dataset is useful for both training and validation, and can assist in developing state-of-the-art solutions for colonoscopy images. The paper also discusses related work, including other available polyp datasets, and highlights the need for open-access datasets for comparable evaluations. The Kvasir-SEG dataset is released as open-source to the multimedia and medical research communities to help evaluate and compare existing and future computer vision methods.