CONVERGE Challenge Dataset: Multimodal Sensing for 6G Wireless Communications (ICASSP 2026)

This record contains the dataset used for the CONVERGE Challenge organized with ICASSP 2026.
The dataset contains synchronized RGB-D video, radio measurements, and annotations for multimodal wireless sensing tasks.

The data are split into training and validation archives. Each split includes scenario indexes, RGB frames, disparity frames, E2 radio measurements, SRS channel measurements, stereo camera calibration, and task annotations.

The dataset supports three challenge tasks: blockage-state prediction, user equipment position estimation, and future SRS channel prediction. Task 1 annotations provide blockage labels with states no, partial, and full. Task 2 annotations provide UE translation coordinates in millimeters. Task 3 uses the same video and radio inputs for future SRS channel prediction.