SIMPL Multi-Aspect MADD: Rapidly Generating Low-Cost Multi-Aspect Military Data for All-Domains at Scale

I. Taylor McKechnie; Leslie M. Collins; Jordan M. Malof

doi:10.1117/12.3053491

SIMPL Multi-Aspect MADD: Rapidly Generating Low-Cost Multi-Aspect Military Data for All-Domains at Scale

I. Taylor McKechnie, Leslie M. Collins, Jordan M. Malof

Computer Science

Duke University

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

In this work we extend an approach known as SIMPL (Synthetic object IMPLantation) to construct a large and diverse synthetic dataset known as MADD (Military All-Domain Dataset). Our extension to SIMPL provides the ability to easily and rapidly generate a massive data set to overcome issues in object recognition algorithm development like: limited data, limited object viewing aspects, data collection infeasibility, or zero-shot problems. In total MADD contains 100 unique targets positioned at a total of 100,000 times in diverse backgrounds. Each target is captured from >50 unique viewing aspects, providing exceptional coverage of what an observing algorithm may see in the wild. We showcase the effectiveness of our method for building or augmenting deep neural networks where there is no real-world data available (Zero-Shot scenarios), and we also demonstrate the dataset's sensitivities using our novel MADD-500 benchmark test set.

Original language	English
Title of host publication	Synthetic Data for Artificial Intelligence and Machine Learning
Subtitle of host publication	Tools, Techniques, and Applications III
Editors	Kimberly E. Manser, Christopher L. Howell, Raghuveer M. Rao, Celso De Melo, Keith F. Prussing
Publisher	SPIE
ISBN (Electronic)	9781510687073
DOIs	https://doi.org/10.1117/12.3053491
State	Published - 2025
Event	Synthetic Data for Artificial Intelligence and Machine Learning: Tools, Techniques, and Applications III 2025 - Orlando, United States Duration: Apr 14 2025 → Apr 17 2025

Publication series

Name	Proceedings of SPIE - The International Society for Optical Engineering
Volume	13459
ISSN (Print)	0277-786X
ISSN (Electronic)	1996-756X

Conference

Conference	Synthetic Data for Artificial Intelligence and Machine Learning: Tools, Techniques, and Applications III 2025
Country/Territory	United States
City	Orlando
Period	04/14/25 → 04/17/25

Keywords

Benchmark Dataset
Deep Neural Networks
Military
Overhead
Synthetic Data
Zero-Shot

Access to Document

10.1117/12.3053491

Cite this

McKechnie, I. T., Collins, L. M., & Malof, J. M. (2025). SIMPL Multi-Aspect MADD: Rapidly Generating Low-Cost Multi-Aspect Military Data for All-Domains at Scale. In K. E. Manser, C. L. Howell, R. M. Rao, C. De Melo, & K. F. Prussing (Eds.), Synthetic Data for Artificial Intelligence and Machine Learning: Tools, Techniques, and Applications III Article 134590F (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 13459). SPIE. https://doi.org/10.1117/12.3053491

McKechnie, I. Taylor ; Collins, Leslie M. ; Malof, Jordan M. / SIMPL Multi-Aspect MADD : Rapidly Generating Low-Cost Multi-Aspect Military Data for All-Domains at Scale. Synthetic Data for Artificial Intelligence and Machine Learning: Tools, Techniques, and Applications III. editor / Kimberly E. Manser ; Christopher L. Howell ; Raghuveer M. Rao ; Celso De Melo ; Keith F. Prussing. SPIE, 2025. (Proceedings of SPIE - The International Society for Optical Engineering).

@inproceedings{1f273486a6fd4a489d0089aaaa979f22,

title = "SIMPL Multi-Aspect MADD: Rapidly Generating Low-Cost Multi-Aspect Military Data for All-Domains at Scale",

abstract = "In this work we extend an approach known as SIMPL (Synthetic object IMPLantation) to construct a large and diverse synthetic dataset known as MADD (Military All-Domain Dataset). Our extension to SIMPL provides the ability to easily and rapidly generate a massive data set to overcome issues in object recognition algorithm development like: limited data, limited object viewing aspects, data collection infeasibility, or zero-shot problems. In total MADD contains 100 unique targets positioned at a total of 100,000 times in diverse backgrounds. Each target is captured from >50 unique viewing aspects, providing exceptional coverage of what an observing algorithm may see in the wild. We showcase the effectiveness of our method for building or augmenting deep neural networks where there is no real-world data available (Zero-Shot scenarios), and we also demonstrate the dataset's sensitivities using our novel MADD-500 benchmark test set.",

keywords = "Benchmark Dataset, Deep Neural Networks, Military, Overhead, Synthetic Data, Zero-Shot",

author = "McKechnie, \{I. Taylor\} and Collins, \{Leslie M.\} and Malof, \{Jordan M.\}",

note = "Publisher Copyright: {\textcopyright} 2025 SPIE. All rights reserved.; Synthetic Data for Artificial Intelligence and Machine Learning: Tools, Techniques, and Applications III 2025 ; Conference date: 14-04-2025 Through 17-04-2025",

year = "2025",

doi = "10.1117/12.3053491",

language = "English",

series = "Proceedings of SPIE - The International Society for Optical Engineering",

publisher = "SPIE",

editor = "Manser, \{Kimberly E.\} and Howell, \{Christopher L.\} and Rao, \{Raghuveer M.\} and \{De Melo\}, Celso and Prussing, \{Keith F.\}",

booktitle = "Synthetic Data for Artificial Intelligence and Machine Learning",

}

McKechnie, IT, Collins, LM & Malof, JM 2025, SIMPL Multi-Aspect MADD: Rapidly Generating Low-Cost Multi-Aspect Military Data for All-Domains at Scale. in KE Manser, CL Howell, RM Rao, C De Melo & KF Prussing (eds), Synthetic Data for Artificial Intelligence and Machine Learning: Tools, Techniques, and Applications III., 134590F, Proceedings of SPIE - The International Society for Optical Engineering, vol. 13459, SPIE, Synthetic Data for Artificial Intelligence and Machine Learning: Tools, Techniques, and Applications III 2025, Orlando, United States, 04/14/25. https://doi.org/10.1117/12.3053491

SIMPL Multi-Aspect MADD: Rapidly Generating Low-Cost Multi-Aspect Military Data for All-Domains at Scale. / McKechnie, I. Taylor; Collins, Leslie M.; Malof, Jordan M.
Synthetic Data for Artificial Intelligence and Machine Learning: Tools, Techniques, and Applications III. ed. / Kimberly E. Manser; Christopher L. Howell; Raghuveer M. Rao; Celso De Melo; Keith F. Prussing. SPIE, 2025. 134590F (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 13459).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - SIMPL Multi-Aspect MADD

T2 - Synthetic Data for Artificial Intelligence and Machine Learning: Tools, Techniques, and Applications III 2025

AU - McKechnie, I. Taylor

AU - Collins, Leslie M.

AU - Malof, Jordan M.

PY - 2025

Y1 - 2025

N2 - In this work we extend an approach known as SIMPL (Synthetic object IMPLantation) to construct a large and diverse synthetic dataset known as MADD (Military All-Domain Dataset). Our extension to SIMPL provides the ability to easily and rapidly generate a massive data set to overcome issues in object recognition algorithm development like: limited data, limited object viewing aspects, data collection infeasibility, or zero-shot problems. In total MADD contains 100 unique targets positioned at a total of 100,000 times in diverse backgrounds. Each target is captured from >50 unique viewing aspects, providing exceptional coverage of what an observing algorithm may see in the wild. We showcase the effectiveness of our method for building or augmenting deep neural networks where there is no real-world data available (Zero-Shot scenarios), and we also demonstrate the dataset's sensitivities using our novel MADD-500 benchmark test set.

AB - In this work we extend an approach known as SIMPL (Synthetic object IMPLantation) to construct a large and diverse synthetic dataset known as MADD (Military All-Domain Dataset). Our extension to SIMPL provides the ability to easily and rapidly generate a massive data set to overcome issues in object recognition algorithm development like: limited data, limited object viewing aspects, data collection infeasibility, or zero-shot problems. In total MADD contains 100 unique targets positioned at a total of 100,000 times in diverse backgrounds. Each target is captured from >50 unique viewing aspects, providing exceptional coverage of what an observing algorithm may see in the wild. We showcase the effectiveness of our method for building or augmenting deep neural networks where there is no real-world data available (Zero-Shot scenarios), and we also demonstrate the dataset's sensitivities using our novel MADD-500 benchmark test set.

KW - Benchmark Dataset

KW - Deep Neural Networks

KW - Military

KW - Overhead

KW - Synthetic Data

KW - Zero-Shot

UR - https://www.scopus.com/pages/publications/105010384440

U2 - 10.1117/12.3053491

DO - 10.1117/12.3053491

M3 - Conference contribution

AN - SCOPUS:105010384440

T3 - Proceedings of SPIE - The International Society for Optical Engineering

BT - Synthetic Data for Artificial Intelligence and Machine Learning

A2 - Manser, Kimberly E.

A2 - Howell, Christopher L.

A2 - Rao, Raghuveer M.

A2 - De Melo, Celso

A2 - Prussing, Keith F.

PB - SPIE

Y2 - 14 April 2025 through 17 April 2025

ER -

McKechnie IT, Collins LM, Malof JM. SIMPL Multi-Aspect MADD: Rapidly Generating Low-Cost Multi-Aspect Military Data for All-Domains at Scale. In Manser KE, Howell CL, Rao RM, De Melo C, Prussing KF, editors, Synthetic Data for Artificial Intelligence and Machine Learning: Tools, Techniques, and Applications III. SPIE. 2025. 134590F. (Proceedings of SPIE - The International Society for Optical Engineering). doi: 10.1117/12.3053491

SIMPL Multi-Aspect MADD: Rapidly Generating Low-Cost Multi-Aspect Military Data for All-Domains at Scale

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this