Journal of the Korea Institute of Military Science and Technology (한국군사과학기술학회지)

The Korea Institute of Military Science and Technology (한국군사과학기술학회)
권호 표지
DOI QR코드

DOI QR Code

Synthetic Image Generation for Military Vehicle Detection

군용물체탐지 연구를 위한 가상 이미지 데이터 생성

  • Se-Yoon Oh (Advanced Defense Science & Technology Research Institute, Agency for Defense Development) ;
  • Hunmin Yang (Advanced Defense Science & Technology Research Institute, Agency for Defense Development)
  • 오세윤 (국방과학연구소 국방첨단과학기술연구원) ;
  • 양훈민 (국방과학연구소 국방첨단과학기술연구원)
  • Received : 2023.04.24
  • Accepted : 2023.11.20
  • Published : 2023.12.05
Download PDF
⟨ Previous Next ⟩

Abstract

This research paper investigates the effectiveness of using computer graphics(CG) based synthetic data for deep learning in military vehicle detection. In particular, we explore the use of synthetic image generation techniques to train deep neural networks for object detection tasks. Our approach involves the generation of a large dataset of synthetic images of military vehicles, which is then used to train a deep learning model. The resulting model is then evaluated on real-world images to measure its effectiveness. Our experimental results show that synthetic training data alone can achieve effective results in object detection. Our findings demonstrate the potential of CG-based synthetic data for deep learning and suggest its value as a tool for training models in a variety of applications, including military vehicle detection.

Keywords

References

  1. S. I. Nikolenko, "Synthetic Data for Deep Learning," Springer Optimization and Its Applications 174, 2021.
  2. A. Tsirikolou, "Synthetic Data for Visual Learning: A Data-Centric Approach," Linkoping Studies in Science and Technology Dissertation No. 2202, Linkoping University, 2022.
  3. Andrej Karpathy, "AI for Full-Self Driving at Tesla," ScaledML Conference, 2020.
  4. Todos os direitos reservados ao Blog do Dercio "VIDEO: cidades de Oregon, nos EUA, tem ceu vermelho devido aos incendios florestais," September 2020.
  5. C. Bressiani Vieira de Rocco, "Synthetic Dataset Creation for Computer Vision Application: Pipeline Proposal," Dissertation, Pontifical Catholic University of PArana Polytechnic School, 2021.
  6. A. Tsirikoglou, G. Eilertsen and L. Unger, "A Survey of Image Synthesis Methods for Visual Machine Learning," Computer Graphics forum, Vol. 39, No. 6, pp. 426-451, 2020. https://doi.org/10.1111/cgf.14047
  7. D. Schraml, "Physically based Synthetic Image Generation for Machine Learning: a Review of Pertinent Literature," Proceedings of SPIE 11144, Sept., 2019.
  8. ImageNet, "Download ImageNet Data," https://image-net.org/download-images, 2020.
  9. Open Images Dataset V4, "Subset with Bounding Boxes(600 classes) and Visual Relationships," https://storage.googleapis.com/openimages/web/download_v4.html, 2019.
  10. NVIDIA, DEEP LEARNING SOFTWARE(SDK), https://www.nvidia.co.kr/content/apac/event/kr/deep-learning-software.
  11. Gartner, "Estimated by Access Ventures," Gartner Report, 2022.
  12. Gartner, "Gartner Emerging Technologies and Trends Impact Radar," 2023.
  13. G. Nam, G. Choi, K. Park, E. Kim and S. Oh, "Synthetic-to-Real Generalization for Object Detection," KIMST Annual Conference Proceedings, pp. 906-907, 2021.
  14. J. Tremblay, et. al., "Training deep networks with synthetic data: Bridging the reality gap by domain randomization," CVPR 2018 Workshop on Autonomous Driving, 2018.
  15. A. Prakash, et. al., "Structured domain randomization: Bridging the reality gap by contextaware synthetic data," arXiv:1810.10093, 2018.
  16. A. Kar, et. al., "Meta-Sim: Learning to Generate Synthetic Datasets," arXiv:1904.11621, 2019.
  17. S. Oh, H. Yang, J. Kim, G. Choi and G. Nam, "Synthetic Image Generation for Deep Learning Projects," KIMST Annual Conference Proceedings, pp. 1305-1306, 2022.
  18. K. Park, H. Lee, H. Yang and S. Oh, "Improving Instance Segmentation using Synthetic Data with Artificial Distractors," 20th International Conference on Control, Automation and Systems(ICCAS 2020), pp. 22-26, 2020.
  19. E. Kim, K. Park, H. Yang and S. Oh, "Training Deep Neural Networks with Synthetic Data for Off-Road Vehicle Detection," 20th International Conference on Control, Automation and Systems (ICCAS 2020), pp. 427-431, 2020.
  20. H. Yang, K. Ryu and S. Oh, "Learning Deep Object Detectors from Synthetic Data," KIMST Annual Conference Proceedings, pp. 699-700, 2019.
  21. S. Oh, K. Ryu and H. Yang, "Optimal Experimental Design for Efficient Image Segmentation," KIMST Annual Conference Proceedings, pp. 697-698, 2019.
  22. H. Yang, K. Ryu and S. Oh, "Improving Deep Learning based Object Detection with Synthetic Data," KIMST Annual Conference Proceedings, pp. 1157-1158, 2019.
  23. H. Yang, K. Ryu and S. Oh, "Cross-Domain Image Translation for Large Shape Transformation using Generative Adversarial Networks," Korea Computer Congress 2018, Vol. 1, pp. 800-801, 2018.
  24. Y. Zhang et al., "CAMOU: Learning Physical Vehicle Camouflages to Adversarial Attack Detectors in the Wild," ICLR, 2019.
  25. T. Wu et al., "Physical Adversarial Attack on Vehicle Detector in the CARLA Simulator," CoRR, abs/2007.16118, 2020.
  26. J. Wang et al., "Dual Attention Suppression Attack: Generate Adversarial Camouflage in Physical World," CVPR, 2021.
  27. D. Wang et al., "FCA: Learning a 3D Full-coverage Vehicle Camouflage for Multi-View Physical Adversarial Attack," AAAI, 2022.
  28. N. Suryanto, Y. Kim, H. Kang, H. T. Larasati, Y. Yun, T.-T.-H. Le, H. Yang, S. Oh, H. Kim, "DTA: Physical Camouflage Attacks using Differentiable Transformation Network," CVPR, 2022.
  29. J. Kim, H. Yang and S. Oh, "Camouflaged Adversarial Patch Attack on Object Detecotr," Journal of the Korea Institute of Military Science and Technology, Vol. 26 No. 1, pp. 44-53, 2023. https://doi.org/10.9766/KIMST.2023.26.1.044
  30. S. Oh, H. Yang and, J. Kim, "Simulation of Physical Adversarial Attack on Object Detection Models," KIMST Annual Conference Proceedings, pp. 589-590, 2022.
  31. H. Yang, S. Oh and, J. Kim, "Physical Adversarial Attacks on Object Detection Models," KIMST Annual Conference Proceedings, pp. 1303-1304, 2022.
  32. H. Yang, S. Oh and, J. Kim, "Threat of 3D Physical Adversarial Attacks on Deep Learning Models," KIMST Annual Conference Proceedings, pp. 595-596, 2022.