EgoWrist
A wrist-mounted perspective focused on close hand-object interaction during real-world tasks.
EgoWrist · EgoExo · Ego
Vanso provides wrist-view, joint first/third-person and first-person multimodal data that captures visual context, hand-object interaction and task progression in the real world.
Discuss a data programA wrist-mounted perspective focused on close hand-object interaction during real-world tasks.
Synchronized first- and third-person views that connect operator intent with complete task context.
First-person multimodal collection with aligned RGB-D, pose, skeleton, annotation and temporal signals.
Ego multimodal specification
The Ego package combines calibrated visual data, motion signals and structured semantic annotations in a model-ready HDF5 delivery.
RGB-D video is delivered with camera intrinsic parameters.
Full-body and both-hand tracking, including every finger joint, with a 4×4 transform matrix and per-frame confidence for each joint.
Six-degree-of-freedom device pose is recorded on the shared timeline.
Chinese and English event-level scene, task and action descriptions, plus hand, object and action labels for every segment.
Actions are segmented by semantic meaning with millisecond-level timing precision.
All modalities are delivered in HDF5 with strictly aligned timestamps.
Representative samples
Collection and production
Our local operations teams can collect first-person demonstrations around a defined task library, sensor stack and delivery schema. Programs are designed to produce usable training signal from the first pilot onward.
Need a tailored dataset, schema or sample package?
contact.vanso@gmail.com