Colonoscopy Video Dataset
A new weakly supervised video benchmark containing a large-scale diverse colonoscopy video dataset that combines several public colonoscopy datasets.
- Colonoscopy Video Dataset: To form a real-world large-scale video polyp detection dataset, we collected colonoscopy videos from two widely used public datasets: Hyper-Kvasir and LDPolypVideo. The new dataset contains 61 normal videos without polyps and 102 abnormal videos with polyps for training, and 30 normal videos and 60 abnormal videos for testing. The videos in the training set have video-level labels and the videos in testing set contain frame-level labels. This dataset contains over one million frames and has diverse polyps with various sizes and shapes, making it one of the largest and most challenging colonoscopy datasets in the field. GitHub Download the Data
