This workshop is organized together with the Amazon’s Scene Understanding Challenge 2023. The dataset of the challenge consisted of birdseye-view images taken between 5m and 25m AGL. Annotations for the semantic segmentation task are fully labeled images, across 16 distinct classes. Annotations for the mono-depth estimation task have been computed with geometric stereo-depth algorithms.

The evaluation of the challenge will be carried according to the instructions provided on the challenge rules on a private test set. There are two separate benchmarks, one for the semantic segmentation task and another for the mono-depth estimation task. In particular the mean IoU will be used as main metric to rank the semantic segmentation leader-board, and the Scale invariant logarithmic error is the main metric used to rank the mono-depth estimation leader-board.

Website

The challenge is hosted through AICrowd.

Dates

December 21st, 2022: Challenge started
April 21st, 2023: Challenge submissions closed
June 30th, 2023: Challenge winners announced

Results!

Winners have been announced on AICrowd:

Semantic segmentation winners: Final Leaderboard Mono-depth estimation winners: Final Leaderboard