Researchers have developed ReferTrack, a novel embodied visual tracking (EVT) system that uses a single forward-facing camera to follow natural language-described targets. This system grounds its decision-making by first identifying the target from bounding boxes and then planning a trajectory. ReferTrack incorporates a temporal memory of past bounding boxes and is enhanced by co-training on a custom Refer-QA dataset. It achieves state-of-the-art performance on the EVT-Bench, outperforming some multi-camera systems, and has demonstrated robust real-world transfer capabilities on legged and humanoid robots. AI
IMPACT This new method could enable more sophisticated robotic navigation and interaction by improving single-camera target tracking capabilities.
RANK_REASON The item describes a new research paper detailing a novel approach to embodied visual tracking. [lever_c_demoted from research: ic=1 ai=1.0]
Read on Hugging Face Daily Papers →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
