Advancements in Human-Robot Interaction and Egocentric Understanding

The field of human-robot interaction and egocentric understanding is moving towards more intuitive and accessible interfaces. Researchers are exploring the use of gaze-guided interaction, wearable devices, and multimodal fusion to enhance the accuracy and robustness of robotic manipulation and object recognition. Notable papers in this area include HRT1, which introduces a novel system for human-to-obot trajectory transfer, and RaycastGrasp, which presents an egocentric gaze-guided robotic manipulation interface. Additionally, papers like Gaze-VLM and GaTector+ are advancing the state-of-the-art in egocentric understanding tasks such as future event prediction and gaze object detection. These innovative approaches are paving the way for more effective and efficient human-robot collaboration.

Sources

HRT1: One-Shot Human-to-Robot Trajectory Transfer for Mobile Manipulation

Sequentially Teaching Sequential Tasks $(ST)^2$: Teaching Robots Long-horizon Manipulation Skills

Gaze-VLM:Bridging Gaze and VLMs through Attention Regularization for Egocentric Understanding

RaycastGrasp: Eye-Gaze Interaction with Wearable Devices for Robotic Manipulation

PixelRefer: A Unified Framework for Spatio-Temporal Object Referring with Arbitrary Granularity

Cross-Enhanced Multimodal Fusion of Eye-Tracking and Facial Features for Alzheimer's Disease Diagnosis

GaTector+: A Unified Head-free Framework for Gaze Object and Gaze Following Prediction

Estimating cognitive biases with attention-aware inverse planning

Built with on top of