Observed arrival · 2026-10-07
TAVIS: A benchmark for robot gaze and active vision
TAVIS is a simulation benchmark for evaluating how robots use movable cameras while learning manipulation tasks from demonstrations.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Robotics researchers studying imitation learning and active vision
- Worth noticing
- Its roughly 2,200 teleoperated demonstrations record head, workspace, and wrist-camera views simultaneously.
Field notes
The demonstrations are described as capturing head, fixed-workspace, and both wrist-camera streams at the same time, allowing comparisons using the same recorded episodes. The page identifies IsaacLab as the simulation basis and links to rollout videos with third-person and robot-camera views. Its abstract reports task-dependent gains from active vision and says action chunking can make head motion less smooth than in human demonstrations.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
◎NicheUnusually specific use
One card from the complete issue