The organizers of the contest have received information about SUNCG which has led us to cease distribution of our DISCOMAN dataset generated from SUNCG. We are trying to find different models for data generation.
Publishing dataset and contest rules 15 July 2019
Opening submission system 15 July 2019
Closing submission system 01 October 2019
Summing up the results 05-06 November 2019
Creating accurate semantic maps of an indoor environment and precise localization in these maps is important for development of the next generation of home robots. The goal of this contest is to explore existing technologies for semantic SLAM, identifying the most suitable solutions and extending the state-of-the-art. We aim at measuring not just localization accuracy, but also the accuracy of mapping component of SLAM.
Odometry : Given an input sequence, estimate corresponding positions and orientations of a robot.
Mapping: Given an input sequence, estimate the top-view occupancy grid for visited parts of the scene.
Segmentation: Given an input sequence, predict the panoptic segmentation labelling for each frame. One can utilize previously seen frames, e.g. create and use an internal scene representation of the house.
Olga Barinova, Samsung AI Center Moscow firstname.lastname@example.org
Pavel Kirsanov, Samsung AI Center Moscow email@example.com
Anton Konushin, Samsung AI Center Moscow
Alexander Vakhitov, Samsung AI Center Moscow
Gonzalo Ferrer, Skolkovo Institute of Science and Technology
Victor Lempitsky, Samsung AI Center Moscow, Skolkovo Institute of Science and Technology