Image classification algorithm performance based on Fλ/d

Jonathan G. Hixson, Brian Teaney, Michael F. Finch, George Nehmetallah, Ronald Driggers

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Scopus citations

Abstract

This paper will take an initial look at the effect of variations in a sensor’s Fλ/d metric value (FLD) on the performance of Yolo_v3 (You Only Look Once) algorithm for object classification. The Yolo_v3 algorithm will initially be trained using static imagery provided in the commonly available Advanced Driver Assist System (ADAS) dataset. Image processing techniques will then be used to degrade image quality of the test data set, simulating detector-limited to optics-limited performance of the imagery. The degraded test set will then be used to evaluate the performance of Yolo_v3 for object classification. Results of Yolo_v3 will be presented for the varying levels of image degradation. An initial summary of the results will be discussed along with recommendations for evaluating an algorithm’s performance using a sensors FLD metric value.

Original languageEnglish (US)
Title of host publicationInfrared Imaging Systems
Subtitle of host publicationDesign, Analysis, Modeling, and Testing XXXV
EditorsDavid P. Haefner, Gerald C. Holst
PublisherSPIE
ISBN (Electronic)9781510674080
DOIs
StatePublished - 2024
EventInfrared Imaging Systems: Design, Analysis, Modeling, and Testing XXXV 2024 - National Harbor, United States
Duration: Apr 23 2024Apr 25 2024

Publication series

NameProceedings of SPIE - The International Society for Optical Engineering
Volume13045
ISSN (Print)0277-786X
ISSN (Electronic)1996-756X

Conference

ConferenceInfrared Imaging Systems: Design, Analysis, Modeling, and Testing XXXV 2024
Country/TerritoryUnited States
CityNational Harbor
Period4/23/244/25/24

Keywords

  • ADAS
  • computer vision
  • convolutional neural network
  • deep learning
  • FLD
  • image identification and classification
  • infrared simulation
  • NV IPM
  • YOLO

ASJC Scopus subject areas

  • Electronic, Optical and Magnetic Materials
  • Condensed Matter Physics
  • Computer Science Applications
  • Applied Mathematics
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'Image classification algorithm performance based on Fλ/d'. Together they form a unique fingerprint.

Cite this