AI
Beyond the Current Scene: Event-Referential Grasping with Active View Selection
Publication: Hugging Face Daily Papers
Author: Not available in collected metadata
Source published: 29/09/2026, 20:00:00 UTC
Briefing created: 03/10/2026, 00:52:00 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 02/10/2026, 04:18:25 UTC.
Briefings
BeyondSCe enables a robot to perform grasping tasks referencing past interactions rather than object names, even when targets are occluded. Using a single wrist-mounted RGB-D camera, the system achieves 76% success rates for initially visible targets and 77% for occluded ones, significantly outperforming baselines. It combines event history with current scene geometry to select optimal camera viewpoints, reducing the mean number of views from 3.35 to 2.20 on four heavy-occlusion scenes.
Key points
- The BeyondSCe system achieves 76% grasp success rates for initially visible targets and 77% for occluded targets using a single wrist-mounted RGB-D camera.
- The system reduces the mean number of camera views from 3.35 to 2.20 on four additional scenes with heavy occlusion compared to an active-perception baseline.
Context and limitations
Results are from real-robot experiments with a single wrist-mounted RGB-D camera; performance on four additional scenes with heavy occlusion is reported but limited by the small sample size.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 03/10/2026, 00:52:00 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeibfr26dgdwzodjpdiwkd2rcqibfcbtxvxvg6z3csb5tkwv5fjwlky