LandNET: A Multi-Modal Fusion Network for Classification

LandNET: A Multi-Modal Fusion Network for Classification

Presenter(s)

Jonathan Paul Schierl

Description

There is a need for classifying land coverage by usage. As these classes are somewhat abstract, this provides a challenge in classifying them and a need for as much information as possible. We propose an architecture capable of classify such scenes, using 2D aerial imagery and 3D point clouds. This is done by fusing the learned feature space of each modality, to be classified with fully connected layers. This method provides a high degree of accuracy for each modality and then learns the benefits of data type, for more accurate classification.

Publication Date

4-22-2021

Project Designation

Graduate Research

Primary Advisor

Theus H. Aspiras

Primary Advisor's Department

Electrical and Computer Engineering

Keywords

Stander Symposium project, School of Engineering

United Nations Sustainable Development Goals

Industry, Innovation, and Infrastructure; Decent Work and Economic Growth

LandNET: A Multi-Modal Fusion Network for Classification

Share

COinS