CVPR 2023 WS: Embodied AI - Robotic Vision Scene Understanding Challenge (RVSU)



~ Please register your interest in the challenge by completing our registration form ~
(these details will help us offer you support and contact prize winners)

The CVPR2023 Embodied AI Workshop of our Robotic Vision Scene Understanding Challenge evaluates how well a robotic vision system can understand the semantic and geometric aspects of its environment. The challenge consists of two distinct tasks: Object-based Semantic SLAM, and Scene Change Detection.

Key features of this challenge include:

Submit your results to EvalAI - Main challenge evaluation servers opening Feb 2023!

Watch the video below to learn more!

Challenge Results

The top two teams this year are

Challenge Task Details

This year, the semantic scene understanding challenge is split into 2 modes, across 2 different semantic scene understanding tasks, for a total of 4 different challenge variations. This year we are not including the passive navigation scene understanding track to place greater emphasis on active embodied agents exploring a scene than in previous iterations of the challenge.

The two different semantic scene understanding tasks are:

  1. Semantic SLAM: Participants use a robot to traverse around the environment, building up an object-based semantic map from the robot’s RGBD sensor observations and odomtry measurements.
  2. Scene change detection (SCD): Participants use a robot to traverse through an environment scene, building up a semantic understanding of the scene. Then the robot is moved to a new start position in the same environment, but with different conditions. Along with a possible change from day to night, the new scene has a number objects added and / or removed. Participants must produce an object-based semantic map describing the changes between the two scenes.

The object-based semantic maps generated by submissions are evaluated against the corresponding ground-truth object-based semantic map (for task 2 this is the map of changes in the second scene, with respect to the first). Please see the BenchBot Evaluation documentation for details on object-based semantic maps submission formats, and further explanation of the two different tasks.

Each task has 2 variations, corresponding to the following modes:

  1. Active control, with ground-truth localisation (AGT): The robot can be controlled by either moving forward a requested distance, or rotating on the spot a requested number of degrees. The task cannot be continued if the robot collides with the environment. Participants receive ground-truth poses for all robot components after each action.
  2. Active control, using dead reckoning (i.e. no localisation) (ADR): The robot can be controlled by either moving forward a requested distance, or rotating on the spot a requested number of degrees. Participants receive poses derived from robot odometry after each action, with localisation error that accumulates over time. Robot movement is also done based on noisy odometry rather than ground truth, providing noise in movement as well as provided poses.

Please see the BenchBot API documentation for full details about what actions and observations are available in each mode, and how to use them with your semantic scene understanding algorithms.

Note that passive mode (as seen in past challenges) is available on the evaluation servers for algorithm development but will not be considered when awarding prizes as part of the main challenge.

How to Participate

The challenge participation page is available on the EvalAI website, and will be open until the challenge closes in May. Please create an account, sign in, & click the “Participate” tab to enter our challenge. Full details on how to participate, the available software framework, and submission requirements are provided on the site.

Participating in the Semantic Scene Understanding Challenge is as simple as the steps below. The BenchBot software stack is designed from the ground up to eliminate as many obstacles as possible, so you can focus on what matters: solving semantic scene understanding problems. A collection of resources, documentation, and examples are also available within the BenchBot ecosystem to support your experience while participating in the challenge.

To participate in our challenge:

  1. Download & install the BenchBot software stack. Use the examples to dive straight in & start playing (benchbot_run –list-examples)
  2. Choose a task to start working on a solution for, using benchbot_run --list-tasks to list supported tasks
  3. Start with the development environments “miniroom” and “house” which include ground-truth maps to aid in your algorithm development
  4. Develop a solution using the benchbot_run, benchbot_submit, & benchbot_eval scripts
  5. Create some results for your solution in the challenge environments using: benchbot_batch -r carter -t <your_task> -E <batch_name> -z -n <your_submission_cmd>
  6. Use the Submit tab at the top of our EvalAI page to submit your results for evaluation

Participants need to provide a short summary of their method at submission time. Each participant is also encouraged to write a short paper (approx 4 pages) outlining the technique which was used in the challenge. These aren’t required at the same time as the initial submission but will be made available to the general public through our challenge website. Please submit summaries and papers by e-mail to

Challenge Prizes

1 RTX A6000 and up to 5 Jetson Nanos (1 for each member) for each of the top 2 winning teams.

Important Dates

Workshop Details

Our Robotic Vision Scene Understanding Challenge is part of the CVPR 2023 Embodied AI workshop. Please check out the workshop website for more details and for information about other robotic vision/embodied AI challenges being run.


Talk to us on Slack or contact us via email at

Organisers, Support, and Acknowledgements

Stay in touch and follow us on Twitter for news and announcements: @robVisChallenge.

David Hall
Commonwealth Scientific and Industrial Research Organisation
Ben Talbot
Queensland University of Technology
Niko Sünderhauf
Queensland University of Technology

The Robotic Vision Challenges organisers have been associated with the Australian Centre for Robotic Vision and QUT Centre for Robotics at Queensland University of Technology (QUT) in Brisbane, Australia. In 2023 David Hall has continued work on the challenges, supported by the Commonwealth Scientific and Industrial Research Organisation (CSIRO) Machine Learning and Artificial Intelligence Future Science Platform (MLAI FSP).

We thank the following supporters of this work: