The field of 3D vision and robotics is rapidly advancing with a focus on improving the accuracy and robustness of various tasks such as pose estimation, scene reconstruction, and object detection. Recent developments have seen a shift towards the use of Gaussian Splatting and other 3D representation techniques to achieve high-fidelity reconstructions and realistic animations. Additionally, there is a growing interest in applying these techniques to real-world applications such as autonomous driving, robotics, and virtual reality. Noteworthy papers in this area include ROPES, which introduces a score-based causal representation learning approach for robotic pose estimation, and DynaPose4D, which presents a novel framework for generating high-quality 4D dynamic content via pose alignment loss. Overall, the field is moving towards more efficient, scalable, and accurate methods for 3D vision and robotics tasks.
Advancements in 3D Vision and Robotics
Sources
ROPES: Robotic Pose Estimation via Score-Based Causal Representation Learning
Robust Point Cloud Reinforcement Learning via PCA-Based Canonicalization
Urban 3D Change Detection Using LiDAR Sensor for HD Map Maintenance and Smart Mobility
Multi-Agent Pose Uncertainty: A Differentiable Rendering Cram\'er-Rao Bound
STG-Avatar: Animatable Human Avatars via Spacetime Gaussian
DynamicTree: Interactive Real Tree Animation via Sparse Voxel Spectrum
A Fully Interpretable Statistical Approach for Roadside LiDAR Background Subtraction
3D Roadway Scene Object Detection with LIDARs in Snowfall Conditions
DynaPose4D: High-Quality 4D Dynamic Content Generation via Pose Alignment Loss
Scaling Up Occupancy-centric Driving Scene Generation: Dataset and Method
UGAE: Unified Geometry and Attribute Enhancement for G-PCC Compressed Point Clouds
EndoWave: Rational-Wavelet 4D Gaussian Splatting for Endoscopic Reconstruction
VR-Drive: Viewpoint-Robust End-to-End Driving with Feed-Forward 3D Gaussian Splatting
Quality-controlled registration of urban MLS point clouds reducing drift effects by adaptive fragmentation
DrivingScene: A Multi-Task Online Feed-Forward 3D Gaussian Splatting Method for Dynamic Driving Scenes
Point-level Uncertainty Evaluation of Mobile Laser Scanning Point Clouds
AtlasGS: Atlanta-world Guided Surface Reconstruction with Implicit Structured Gaussians
EA3D: Online Open-World 3D Object Extraction from Streaming Videos
$D^2GS$: Dense Depth Regularization for LiDAR-free Urban Scene Reconstruction
U-CAN: Unsupervised Point Cloud Denoising with Consistency-Aware Noise2Noise Matching
JOGS: Joint Optimization of Pose Estimation and 3D Gaussian Splatting
The Impact and Outlook of 3D Gaussian Splatting