Releasing Awesome UMI, a comprehensive catalogue of UMI devices for robot imitation learning data
commissure Inc. has published Awesome-UMI, a catalogue site for comparing UMI devices — data collection devices for imitation learning — side by side.
Imitation learning depends on demonstration data recorded by people. Since UMI (Universal Manipulation Interface), derivative devices have appeared rapidly: versions with tactile sensing, multi-finger support, wearable form factors and more. Yet there was no single place to compare them. Awesome-UMI answers this with a catalogue that can be filtered by form factor, presence of tactile sensing and other criteria.
Thirty-nine devices are registered so far (including 18 handheld devices and many with tactile sensing). The data is published as JSON under CC BY 4.0, and additions and corrections from the community are welcome.
See also this article for details.
https://note.com/arata_horie/n/ne920fe5997b0