arXivSub Start free trial

ICRA 2025 Papers — Page 17

IEEE International Conference on Robotics and Automation · 1604 papers

ZeroMimic: Distilling Robotic Manipulation Skills from Web Videos

Junyao Shi, Dinesh Jayaraman

Knowledge DistillationRobotic IntelligenceVideo

🎯 What it does: Based on large-scale human network videos, ZeroMimic generates directly deployable image-goal conditioned manipulation policies through knowledge distillation, covering multiple tasks such as opening, closing, pouring, grasping, cutting, and stirring.

ZeroSCD: Zero-Shot Street Scene Change Detection

S. S. Kannan, Byung-Cheol Min

SegmentationAutonomous DrivingConvolutional Neural NetworkImageBenchmark

🎯 What it does: Propose a zero-shot street scene change detection framework called ZeroSCD, which detects changes between two images of the same scene captured at different times by leveraging features and outputs from pre-trained location recognition models and semantic segmentation models without any training.

Zippy: The Smallest Power-Autonomous Bipedal Robot

Steven Man, Sarah Bergbreiter

Robotic Intelligence

🎯 What it does: Designed and controlled Zippy, a self-contained bipedal walking robot only 3.6 cm tall.

ZSORN: Language-Driven Object-Centric Zero-Shot Object Retrieval and Navigation

Tianrui Guan, Dinesh Manocha

RetrievalAutonomous DrivingLarge Language ModelSupervised Fine-TuningPrompt EngineeringVision Language ModelImageText

🎯 What it does: Proposed a language-driven, object-oriented image representation method called ZSORN for object retrieval and navigation tasks in complex scenes.