Chimney or squirrel? Clearing up a robot's fuzzy memory by comparing images from different angles
Chimney or squirrel? Clearing up a robot's fuzzy memory by comparing images from different angles
MIT researchers have developed UQ-DAAAM, a new memory system for robots that identifies and corrects mislabeled objects by comparing descriptions from multiple viewpoints. The system quantifies uncertainty in vision-language model outputs, selects additional camera views to reduce ambiguity, and fuses captions into more accurate object descriptions. In tests, the method resolved ambiguities for about 92% of objects and improved description accuracy by 5% compared to baseline methods. The research is presented as a preprint on arXiv and will be discussed at the Conference on Neural Information Processing Systems.
Full article
Article text is not available for this item. Read it at the original source.
Read full article at sourcemsn.com · Opens in a new tab; source language may differ.
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
Reported accuracy
5%
View original evidence
them into a single object description. To boost accuracy, the LLM puts more weight on new captions and those with less ambiguity. This refinement step can resolve ambiguities for about 92% of the objects in the robot's environment while requiring only two additional captions per object. It increased the accuracy of object descriptions by 5% compared to the state of the art. UQ-DAAAM also provided more accurate answe
Open source S14additional captions per object
2
View original evidence
them into a single object description. To boost accuracy, the LLM puts more weight on new captions and those with less ambiguity. This refinement step can resolve ambiguities for about 92% of the objects in the robot's environment while requiring only two additional captions per object. It increased the accuracy of object descriptions by 5% compared to the state of the art. UQ-DAAAM also provided more accurate answe
Open source S14
- paper: not_reported
What remains unknown
Not established in the collected evidence: Environment, Control, Data origin.
Reported performance applies to the described task. It does not establish general autonomy or deployment readiness.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
them into a single object description. To boost accuracy, the LLM puts more weight on new captions and those with less ambiguity. This refinement step can resolve ambiguities for about 92% of the objects in the robot's environment while requiring only two additional captions per object. It increased the accuracy of object descriptions by 5% compared to the state of the art. UQ-DAAAM also provided more accurate answe
Open source S14
rs than baseline methods to questions about the objects it observed, where it saw them and when it last encountered them. "Performing this uncertainty quantification procedure across the entire map the robot generates allows it to improve its understanding of what it has observed," Gorlo says. Next, the researchers want to use UQ-DAAAM to guide different types of robots, like mobile robotic manipulators, and apply th
Open source S15
Source：Industrial robotics — News discovery · msn.com