Graph Convolutional Network for 3D Object Pose Estimation in a Point Cloud

Article Computer Science, Artificial Intelligence

DeepGCNs: Making GCNs Go as Deep as CNNs

Guohao Li et al.

Summary: Convolutional neural networks (CNNs) have been successful in computer vision tasks, but are not suited for non-Euclidean data. Graph Convolutional Networks (GCNs) can handle non-Euclidean data input. This work introduces concepts from CNNs to GCNs and successfully trains very deep GCNs, achieving promising performance in various datasets and tasks.

IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE (2023)

Add to Collection

Article Chemistry, Multidisciplinary

Point-Graph Neural Network Based Novel Visual Positioning System for Indoor Navigation

Tae-Won Jung et al.

Summary: This study proposed a novel visual positioning system using point cloud and RGB characteristic information to accurately acquire three-dimensional indoor space. By matching the pose information of the object estimated by an improved point-graph neural network with the pose information label in predefined voxel units of a voxel database object, the user's position can be determined. Results show that high positioning accuracy and direction estimation can be efficiently achieved in indoor environments using this method.

APPLIED SCIENCES-BASEL (2021)

Add to Collection

Review Chemistry, Multidisciplinary

Object Detection, Distributed Cloud Computing and Parallelization Techniques for Autonomous Driving Systems

Edgar Cortes Gallardo Medina et al.

Summary: This paper reviews the latest machine vision algorithms used in assisted driving and autonomous vehicle development, exploring concepts like distributed systems and parallelization. Autonomous vehicles are a necessary trend towards building smart cities of the future, but efforts are still needed to address challenges like real-time computation and hardware capabilities. Techniques such as convolutional neural networks and recurrent neural networks are effective for initial detection and path planning tasks, but more work is needed to implement cloud computing for reduced computational time.

APPLIED SCIENCES-BASEL (2021)

Add to Collection

Article Geography, Physical