Event

Shanghai AI Lab and others release EgoExplorer wearable capture system and 100-hour dataset

1 reports · 1 sources

A research team from Shanghai AI Lab, Zhejiang University, and Shanghai Jiao Tong University proposed a wearable capture system called EgoExplorer, and used it to compile a 100-hour human activity dataset for robot training. The system uses a head-mounted GoPro HERO13, 5 VIVE Ultimate Trackers, and MANUS Metagloves or Wuji gloves to synchronously record first-person video, full-body motion, finger movements, and tactile observations. The team also proposed a Calibrate-Complete-Constrain motion reconstruction pipeline. In preliminary learning experiments, generating video, motion, and tactile data simultaneously outperformed generating them separately. However, the research was tested only in simulated environments: the simulated robot followed two motion sequences without falling, and it did not demonstrate precise object interaction or good performance on new tasks, nor was it deployed and tested on a physical robot.