A5026778364- Robot Manipulation and Learning
- Soft Robotics and Applications
- 3D Shape Modeling and Analysis
- Remote Sensing and LiDAR Applications
- 3D Surveying and Cultural Heritage
- Reinforcement Learning in Robotics
- Infrastructure Maintenance and Monitoring
- Robotic Mechanisms and Dynamics
- Robotics and Sensor-Based Localization
- Industrial Vision Systems and Defect Detection
- Image and Object Detection Techniques
- Human Pose and Action Recognition
Shanghai Jiao Tong University
2023-2024
Point cloud classification and segmentation are crucial tasks for point processing have wide range of applications, such as autonomous driving robot grasping. Some pioneering methods, including PointNet, VoxNet, DGCNN, etc., made substantial advancements. However, most these methods overlook the geometric relationships between points at large distances from different perspectives within cloud. This oversight constrains feature extraction capabilities consequently limits any further...
6-DoF object-agnostic grasping in unstructured environments is a critical yet challenging task robotics. Most current works use non-optimized approaches to sample grasp locations and learn spatial features without concerning the task. This paper proposes GraNet, graph-based pose generation framework that translates point cloud scene into multi-level graphs propagates through graph neural networks. By building at level, object GraNet enhances feature embedding multiple scales while...
Automatic detection for screws and screw holes is crucial the automatic assembly disassembly of on production line. The mainstream schemes mainly include vision-based methods, deep learning based methods in an end-to-end fashion, combinations two. In this paper, we suggest that semantic segmentation models combining with post processing can boost performance positioning identification mobile phone PCB. our experiment, model correctly detected all stable condition; vibrating conditions,...
Point cloud classification and segmentation are crucial tasks for point processing have wide range of applications, such as autonomous driving robot grasping. Some pioneering methods, including PointNet, VoxNet, DGCNN, etc., made substantial advancements. However, most these methods don't consider the large-distance geometric relationships among points in different perspectives within cloud, which limits features extraction leads to accuracy cannot be further improved. To address this issue,...
Abstract Estimating the orientation and position of objects is a crucial step in robotic bin-picking tasks. The challenge lies fact that, real-world scenarios, diverse array often randomly stacked, resulting significant occlusion. This study introduces an innovative approach aimed at predicting 6D poses by processing point clouds through two-stage neural network. In initial stage, network for scenes with low-textured environments designed. Its purpose to perform instance segmentation provide...
6-DoF object-agnostic grasping in unstructured environments is a critical yet challenging task robotics. Most current works use non-optimized approaches to sample grasp locations and learn spatial features without concerning the task. This paper proposes GraNet, graph-based pose generation framework that translates point cloud scene into multi-level graphs propagates through graph neural networks. By building at level, object GraNet enhances feature embedding multiple scales while...