Determining the utility of a new metric using an object detector

Igor Semyonov; J. L. Pezzaniti; Frank P. Pantuso; David Chenault; Jerome A. Zadnik; Collin J. Bright

doi:10.1117/12.2622929

30 May 2022 Determining the utility of a new metric using an object detector

Igor Semyonov, J. L. Pezzaniti, Frank P. Pantuso, David Chenault, Jerome A. Zadnik, Collin J. Bright

Proceedings Volume 12116, Chemical, Biological, Radiological, Nuclear, and Explosives (CBRNE) Sensing XXIII; 121160Z (2022) https://doi.org/10.1117/12.2622929
Event: SPIE Defense + Commercial Sensing, 2022, Orlando, Florida, United States

Abstract

We investigated the utility of target visibility and dissimilarity metrics proposed by Pezzaniti et al.? Metric utility was determined by looking for correlation between the metrics and ML object detection performance. To do this, thousands of synthetic image sets were generated with varied seed parameters, such as target mean, background mean, target contrast and background contrast. The visibility and dissimilarity metrics for the synthetic images were calculated. YOLOv5? was used to detect simulated military threats in the synthetic data. We will present correlations to determine the significance of the visibility and dissimilarity metrics as it relates to detection performance.

Conference Presentation

Citation Download Citation

Igor Semyonov, J. L. Pezzaniti, Frank P. Pantuso, David Chenault, Jerome A. Zadnik, and Collin J. Bright "Determining the utility of a new metric using an object detector", Proc. SPIE 12116, Chemical, Biological, Radiological, Nuclear, and Explosives (CBRNE) Sensing XXIII, 121160Z (30 May 2022); https://doi.org/10.1117/12.2622929

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available