Our U-net wins two Challenges at ISBI 2015

(April 2015) At the International Symposium on Biomedical Imaging (April 16-19 2015 in New York) we won two challenges with our new deep convolutional network "u-net".

Winner of the Dental X-Ray Image Segmentation Challenge

In the Grand Challange for Computer-Automated Detection of Caries in Bitewing Radiography we trained our u-net to segment dental x-ray images, and outperformed the second best method by a large margin. We won the offline competition before the conference and the online competition at ISBI 2015.

We are among the 4 Winners of the Cell Tracking Challenge (Third Edition)

The Cell Tracking Challenge compares the performance of segmentation and tracking algorithms on a set of 13 very different microscopic time series of living cells. We trained our u-net for segmentation of the two most challenging 2D datasets (U373 cells in phase constrast microscopy and HeLa cells in DIC microscopy) and outperformed the other approaches by a large margin. We also applied the u-net to 4 further 2D data sets (without any adjustment of the parameters) and ranked among the top-three on 3 of these 4 data sets. Due to the large diversity of the applications and the possibility to submit results only for a subset of applications, the challenge organizers selected four winners without further ranking (from 14 participating teams) based on the number of times the groups were listed among the TOP3 groups. We are among these winners.

Further Information