Fraunhofer HHI wins second place at EndoVis 2019 challenge “SCARED”

November 19, 2019

At the “Stereo Correspondence and Reconstruction of Endoscopic Data(SCARED) challenge in Shenzhen, China, Jean-Claude Rosenthal, Research Associate in the Vision and Imaging Technologies Department at Fraunhofer HHI, and his colleagues Zhenglei Hu, Niklas Gard and Peter Eisert won second place in the category „lowest mean error“. The challenge was organized and assessed by Intuitive Surgical. It is part of the Endoscopic Vision (EndoVis) Challenge 2019, endorsed by the International Society for Computer Aided Surgery (ISCAS).

The “SCARED” challenge has been hosted as a satellite event during the renowned conference for Medical Image Computing and Computer Assisted Intervention Society (MICCAI) in Shenzhen, China. The challenge focuses on depth estimation from stereo endoscopic data in the field of minimally invasive surgery. Depth estimation as well as 3D scene reconstruction are important components of many endoscopic navigation and Augmented Reality surgical guidance systems.

Fraunhofer HHI researcher Jean-Claude Rosenthal and his colleagues from the Vision and Imaging Technologies Department Zhenglei Hu, Niklas Gard and Peter Eisert contributed to that challenge by implementing a hybrid approach of advanced handcrafted stereo computer vision algorithms and a DNN for depth completion at object boundaries. The challenge dataset was downloaded 54 times with 10 teams submitting their results to the challenge platform. Fraunhofer HHI results are just 0.1mm behind 1st place in terms of mean absolute error over all challenge data sets.