Sensitivity Analysis of Convolutional Neural Network

Author Information

Annaelise SwansonFollow

Is this project an undergraduate, graduate, or faculty project?

Undergraduate

Project Type

individual

Campus

Daytona Beach

Authors' Class Standing

Annaelise Swanson, Junior

Lead Presenter's Name

Annaelise Swanson

Lead Presenter's College

DB College of Arts and Sciences

Faculty Mentor Name

Mihhail Berezovski

Abstract

This study aims to develop a Convolutional Neural Network (CNN) that is capable of accurately identifying the spectra value of materials in photographs. Due to their capacity for capturing spatial features and patterns, CNNs are very useful for the analysis of image data. By introducing noise to the images, we will be able to gauge how well the CNN performs by evaluating its sensitivity and analyzing its error loss consistency. Insights on the CNN's efficiency in evaluating picture data and its potential for use in noisy image data in the real world will be provided by the study's findings. This study could help strengthen CNNs so they can handle noisy picture data more successfully. Overall, this study highlights the importance of measuring the sensitivity of CNNs to evaluate their performance in image data analysis, particularly when dealing with noisy image data.

Did this research project receive funding support (Spark, SURF, Research Abroad, Student Internal Grants, Collaborative, Climbing, or Ignite Grants) from the Office of Undergraduate Research?

Yes, SURF

Share

COinS
 

Sensitivity Analysis of Convolutional Neural Network

This study aims to develop a Convolutional Neural Network (CNN) that is capable of accurately identifying the spectra value of materials in photographs. Due to their capacity for capturing spatial features and patterns, CNNs are very useful for the analysis of image data. By introducing noise to the images, we will be able to gauge how well the CNN performs by evaluating its sensitivity and analyzing its error loss consistency. Insights on the CNN's efficiency in evaluating picture data and its potential for use in noisy image data in the real world will be provided by the study's findings. This study could help strengthen CNNs so they can handle noisy picture data more successfully. Overall, this study highlights the importance of measuring the sensitivity of CNNs to evaluate their performance in image data analysis, particularly when dealing with noisy image data.