Robotic Vision Scene Understanding Challenge Call for Papers

 Call for Participants: Robotic Vision Scene Understanding Challenge

The Australian Centre for Robotic Vision is pleased to announce a new
robotic vision challenge on scene understanding. In our challenge,
participants will control a robotic agent using simple OpenAI
Gym-style controls within a virtual environment to map out the cuboid
locations of objects in 3D space. A cash prize of $2,500 USD will be
split among high-performing participants in our challenge and they
will also receive the opportunity to run their scene understanding
algorithms, with no modifications, on a real robotic platform.

Challenge Link:
 Important Dates

    Final submissions to EvalAI Due - 2nd September 2020
    Accompanying Paper submissions - 2nd October 2020
    Result notifications - 16th October 2020

The Robotic Vision Scene Understanding Challenge evaluates how well a
robotic vision system can understand the semantic and geometric
aspects of its environment. There are two tasks in this challenge:
Object-based Semantic Mapping/SLAM, and Scene Change Detection.

    Semantic SLAM: Participants use a robot to traverse around the
    environment, building up an object-based semantic map.

    Scene change detection (SCD): Participants use a robot to traverse
    through two different instances of an environment. Between
    instances some objects are added or removed and participants must
    produce an object-based semantic map describing the changes
    between scenes.

Each task has three difficulty levels with lowest difficulty level
requiring no active navigation or localization from the participant,
the next level requiring navigation but not localization, and the
highest level requiring both as the simulation becomes more akin to a
real robot.

Other key features of the challenge include:

    BenchBot a complete software stack for running semantic scene
    understanding algorithms

    The BenchBot API allowing simple interfacing with robots,
    supporting OpenAI Gym-style approaches

    Running algorithms in realistic 3D simulation powered by Nvidia's
    Isaac simulator, and on real robots, with only a few lines of
    Python code

    Easy-to-use-scripts for running simulated environments, executing
    code on a simulated robot, evaluating semantic scene understanding
    results, and automating code execution across multiple

    Opportunities for the best teams to execute their code on a real
    robot in our lab

More Information:
Challenge server:
BenchBot software stack:
Challenge overview video:

Contact Us:
Slack Workspace:
Twitter: @robVisChallenge