The HEART-MET cascade campaign Dataset
This dataset relates to the METRICS HEART-MET Assistive Robot Challenge at IROS 2022, which will be conducted entirely online based on datasets.
Three challenges are included in this competition: gesture recognition, activity recognition, and handover failure detection.
The gesture and activity recognition datasets comprise of short video clips of persons performing different gestures and activities in domestic environments, which need to be recognized. The datasets also contain out-of-distribution (OOD) samples, including corrupted and irrelevant videos - therefore the task requires detection of OOD samples as well.
The handover failure detection dataset consists of video and force-torque data from two robot platforms performing robot-to-human and human-to-robot handovers. The task is to detect the outcome of the handover action, including several failure modes which are induced during execution.