Year: 2023
Type: book-chapter
Abstract: The recognition of 3D objects that are represented as point cloud has become a very vital topic for research in the field of Computer Science.The processing of 3D point clouds has procured a lot of attention from researchers in the modern era.However, as point clouds have a complex representation, it is still a challenging task to capture 3D objects using LiDAR (Light Detection and Ranging) devices.The goal of our work is to identify an object that is represented as a point cloud accurately using 3D Deep Learning.The scope of the project is very vast, and its applications include AR/VR (augmented/virtual reality), self-driving vehicles, robotics, etc.The work aims to classify the 3D objects and also to identify various integrated parts of the object by performing segmentation by accessing the 3d point clouds directly on modelnet10 and shapenet for classification and segmentation respectively.In classification, the captured object is classified into several classes such as a chair, table, nightstand, sofa, bed, etc.Using the most suitable deep learning model we perform classification on the point clouds obtained from the ModelNet10 dataset and segment the objects into parts using shapenet dataset.Also, experimentation was done on different learning rate parameters for better results.
Authors: D. Jyothsna, G. Ramesh Chandra
Language: English
Subfield: Computational Mechanics
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
Award: G1702879744
OpenAlex ID: W4390074414