CaDIS Dataset

CaDIS: a Cataract Dataset for Image Segmentation is a dataset for semantic segmentation created by Digital Surgery Ltd. on top of the CATARACTS dataset. CaDIS consists of 4670 images sampled from the 25 videos on CATARACTS' training set. Each pixel in each image is labeled with its respective instrument or anatomical class from a set of 36 identified classes. More details about the dataset could be found in the paper (https://arxiv.org/pdf/1906.11586.pdf).

Who is Digital Surgery?

Digital Surgery is a health tech company, based in London, UK, shaping the future of surgery through the convergence of surgical expertise and technology. The Innovation team is working on bridging the gap between Artificial intelligence and the OR.

Why are we releasing CaDIS to the public?

We believe releasing a semantic dataset will encourage the computer vision community to push surgical research further.

Data

The dataset consists of 4670 images in total and includes 36 classes: 29 surgical instrument classes, 4 anatomy classes and 3 classes of other object appearing in the scene. The following table gives an overview of the classes included in the dataset with their respective class ID per category.  The training, validation and test sets contain 3550, 534 (Videos 5, 7 and 16) and 586 (Videos 2, 12 and 22) images respectively.

Download the dataset

Uploading - Link coming up soon