The field of 3D perception and autonomous systems is rapidly advancing, with a focus on improving the accuracy and robustness of 3D object detection, scene understanding, and navigation. Recent developments include the use of weakly supervised learning methods, such as Weak Cube R-CNN, which can predict 3D objects using only 2D bounding box annotations. Additionally, techniques like Mono3R and MS-Occ are being explored to enhance 3D reconstruction and occupancy prediction. The integration of LiDAR and camera data is also becoming increasingly important, with methods like RCAlign and MS-Occ demonstrating state-of-the-art performance in 3D object detection and semantic occupancy prediction. Furthermore, researchers are investigating the use of graph-based methods, such as Graph2Nav, to improve navigation and object-relation graph generation in autonomous systems. Noteworthy papers in this area include Weak Cube R-CNN, which achieves increased performance in accuracy compared to an annotation time equalized baseline, and Mono3R, which substantially enhances the robustness of multi-view reconstruction systems. Overall, these advancements have the potential to significantly improve the performance and safety of autonomous systems in various applications, including robotics, autonomous driving, and augmented reality.
Advancements in 3D Perception and Autonomous Systems
Sources
Weak Cube R-CNN: Weakly Supervised 3D Detection using only 2D Bounding Boxes
Mono3R: Exploiting Monocular Cues for Geometric 3D Reconstruction
Testing the Fault-Tolerance of Multi-Sensor Fusion Perception in Autonomous Driving Systems
WeatherGen: A Unified Diverse Weather Generator for LiDAR Point Clouds via Spider Mamba Diffusion
Leveraging Automatic CAD Annotations for Supervised Learning in 3D Scene Understanding
HAECcity: Open-Vocabulary Scene Understanding of City-Scale Point Clouds with Superpoint Graph Clustering
LMPOcc: 3D Semantic Occupancy Prediction Utilizing Long-Term Memory Prior from Historical Traversals
Lightweight LiDAR-Camera 3D Dynamic Object Detection and Multi-Class Trajectory Prediction
RefComp: A Reference-guided Unified Framework for Unpaired Point Cloud Completion
V2P Collision Warnings for Distracted Pedestrians: A Comparative Study with Traditional Auditory Alerts
VoxCity: A Seamless Framework for Open Geospatial Data Integration, Grid-Based Semantic 3D City Model Generation, and Urban Environment Simulation
Occlusion-Ordered Semantic Instance Segmentation
HFBRI-MAE: Handcrafted Feature Based Rotation-Invariant Masked Autoencoder for 3D Point Cloud Analysis
Locate 3D: Real-World Object Localization via Self-Supervised Learning in 3D
Exploring Modality Guidance to Enhance VFM-based Feature Fusion for UDA in 3D Semantic Segmentation
SG-Reg: Generalizable and Efficient Scene Graph Registration
ApexNav: An Adaptive Exploration Strategy for Zero-Shot Object Navigation with Target-centric Semantic Fusion
RoboOcc: Enhancing the Geometric and Semantic Scene Understanding for Robots
VM-BHINet:Vision Mamba Bimanual Hand Interaction Network for 3D Interacting Hand Mesh Recovery From a Single RGB Image
NVSMask3D: Hard Visual Prompting with Camera Pose Interpolation for 3D Open Vocabulary Instance Segmentation
An Iterative Task-Driven Framework for Resilient LiDAR Place Recognition in Adverse Weather
ScanEdit: Hierarchically-Guided Functional 3D Scan Editing
VistaDepth: Frequency Modulation With Bias Reweighting For Enhanced Long-Range Depth Estimation
Robust and Real-time Surface Normal Estimation from Stereo Disparities using Affine Transformations
Locating and Mitigating Gradient Conflicts in Point Cloud Domain Adaptation via Saliency Map Skewness
MS-Occ: Multi-Stage LiDAR-Camera Fusion for 3D Semantic Occupancy Prediction
Understanding the Role of Covariates in Numerical Reconstructions of Real-World Vehicle-to-Pedestrian Collisions
ForesightNav: Learning Scene Imagination for Efficient Exploration
Context-Awareness and Interpretability of Rare Occurrences for Discovery and Formalization of Critical Failure Modes
MonoTher-Depth: Enhancing Thermal Depth Estimation via Confidence-Aware Distillation
Measuring Uncertainty in Shape Completion to Improve Grasp Quality
PCF-Grasp: Converting Point Completion to Geometry Feature to Enhance 6-DoF Grasp
Revisiting Radar Camera Alignment by Contrastive Learning for 3D Object Detection
DPGP: A Hybrid 2D-3D Dual Path Potential Ghost Probe Zone Prediction Framework for Safe Autonomous Driving
Gaussian Splatting is an Effective Data Generator for 3D Object Detection
Graph2Nav: 3D Object-Relation Graph Generation to Robot Navigation
Procedural Dataset Generation for Zero-Shot Stereo Matching
MobileCity: An Efficient Framework for Large-Scale Urban Behavior Simulation
Scene-Aware Location Modeling for Data Augmentation in Automotive Object Detection
AUTHENTICATION: Identifying Rare Failure Modes in Autonomous Vehicle Perception Systems using Adversarially Guided Diffusion Models
Highly Accurate and Diverse Traffic Data: The DeepScenario Open 3D Dataset
S2S-Net: Addressing the Domain Gap of Heterogeneous Sensor Systems in LiDAR-Based Collective Perception
Predict-Optimize-Distill: A Self-Improving Cycle for 4D Object Understanding
Flying through cluttered and dynamic environments with LiDAR
Improving Open-World Object Localization by Discovering Background
DiMeR: Disentangled Mesh Reconstruction Model
PICO: Reconstructing 3D People In Contact with Objects
Gripper Keypose and Object Pointflow as Interfaces for Bimanual Robotic Manipulation
The Fourth Monocular Depth Estimation Challenge
LiDPM: Rethinking Point Diffusion for Lidar Scene Completion