We introduce GPOcc, a novel framework that leverages generalizable Geometry Priors (GPs) for Occupancy prediction with sparse Gaussian rendering.O ur method surpasses the previous state of the art by 9.99 mIoU and 8.24 IoU in the monocular setting, while running at 2.65X FPS.
We introduce LegoOcc, which leverages Language-embedded Gaussians as a fine-grained 3D spatial intermediate representation for monocular Open-vocabulary Occupancy prediction, enabling embodied agents to reason about arbitrary objects beyond closed taxonomies.
PeakLab@HKUST-GZ will present #ICRA2026 — advancing large‑scale thermal SLAM.
🚨 LST-SLAM: A Stereo Thermal SLAM System for Kilometer‑Scale Dynamic Environments
📅 Wednesday, June 3rd | 15:00–16:30 | WeI2I.81
Stop by and chat with us! 🤖
arxiv.org/abs/2602.20925#Robotics#SLAM
PeakLab@HKUST-GZ will present at #ICRA2026 — advancing safe navigation
🚀 SaferPath: Hierarchical Visual Navigation with Learned Guidance & Safety‑Constrained Control
📅 Thu June 4, 9:00–10:30 ThI1I.272
Come chat with us if you’re around!
arxiv.org/abs/2603.01898#Robotics#SLAM
Excited to be accepted into RSS Pioneer 2020. I will introduce our work on "Learning Methods for Robust Localization" this Saturday (virtually).
This 3-minutes video summarizes my research during my Ph.D. study: youtu.be/RVyifQ5TUpE#RSS2020#Robotics
Check our survey paper on "Deep Learning for Localization and Mapping: Towards the Age of Spatial Machine Intelligence": arxiv.org/abs/2006.12567.
We extensively cover the topics from odometry estimation, mapping, to global localization and SLAM. #Robotics#ComputerVision
We hope our survey can connect emerging works from robotics, computer vision and machine learning communities, and serve as a guide for future researchers to apply deep learning to tackle localization and mapping problems.
Project Website: github.com/changhao-chen/dee…
We show how learned sensor fusion strategies can improve accuracy & robustness in deep VIO when dealing with noisy/corrupted data, while adding interpretability.
Selective Sensor Fusion for Visual-Inertial Odometry arxiv.org/abs/1903.01534
Poster 208 Thu 20, 10:00 am #CVPR2019