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

A Study on the Explainability of Inception Network-Derived Image Classification AI Using National Defense Data

국방 데이터를 활용한 인셉션 네트워크 파생 이미지 분류 AI의 설명 가능성 연구

  • Kangun Cho (Technology & Information Security Center - Information System & Common Technology Division, Agency for Defense Development)
  • 조강운 (국방과학연구소 기술정보보안센터 정보화기술실)
  • Received : 2023.07.20
  • Accepted : 2024.02.05
  • Published : 2024.04.05
Download PDF
⟨ Previous Next ⟩

Abstract

In the last 10 years, AI has made rapid progress, and image classification, in particular, are showing excellent performance based on deep learning. Nevertheless, due to the nature of deep learning represented by a black box, it is difficult to actually use it in critical decision-making situations such as national defense, autonomous driving, medical care, and finance due to the lack of explainability of judgement results. In order to overcome these limitations, in this study, a model description algorithm capable of local interpretation was applied to the inception network-derived AI to analyze what grounds they made when classifying national defense data. Specifically, we conduct a comparative analysis of explainability based on confidence values by performing LIME analysis from the Inception v2_resnet model and verify the similarity between human interpretations and LIME explanations. Furthermore, by comparing the LIME explanation results through the Top1 output results for Inception v3, Inception v2_resnet, and Xception models, we confirm the feasibility of comparing the efficiency and availability of deep learning networks using XAI.

Keywords

References

  1. KRIZHEVSKY, Alex; SUTSKEVER, Ilya; HINTON, Geoffrey E, "Imagenet Classification with Deep Convolutional Neural Networks," Communications of the ACM, 60.6: pp. 84-90, 2017. https://doi.org/10.1145/3065386
  2. GOLDSTEIN, Alex, et al., "Peeking Insde the Black Box: Visualizaing Statistical Learning with Plots of Individual Conditional Expectation," Journal of Computational and Graphical Statistics, 24.1: pp. 44-65, 2015. https://doi.org/10.1080/10618600.2014.907095
  3. LUNDBERG, Scott M.; LEE, Su-In, "A Unified Approach to Interpreting Model Predictions," Advances in Neural Information Processing Systems, 30, 2017.
  4. Ribeiro, Marco Tulio, Sameer Singh, and Carlos Guestrin, ""Why Should I Trust You?" Explaining the Predictions of any Classifier," Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016.
  5. SZEGEDY, Christian, et al., "Rethinking the Inception Architecture for Computer Vision," In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 2818-2826, 2016.
  6. SZEGEDY, Christian, et al., "Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning," arXiv preprint arXiv:1602.07261, 2016.
  7. CHOLLET, Francois, "Xception: Deep Learning with Depthwise Separable Convolutions," In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1251-1258, 2017.
  8. SZEGEDY, Christian, et al., "Going Deeper with Convolutions," In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1-9, 2015.
  9. RUDER, Sebastian, "An Overview of Gradient Descent Optimization Algorithms," arXiv preprint arXiv:1609.04747, 2016.
  10. HE, Kaiming, et al., "Deep Residual Learning for Image Recognition," In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 770-778, 2016.
  11. SELVARAJU, Ramprasaath R., et al., "Grad-Cam: Visual Explanations from Deep Neural Networks Via Gradient-based Localization." In: Proeedings of the IEEE International Conference on Computer Vision, pp. 618-626, 2017.
  12. BINDER, Alexandar, et al., "Layer-Wise Relevance Propagation for Neural Networks with Local Renormalization Layers," In: Artificial Neural Networks and Machine Learning-ICANN 2016: 25th International Conference on Artificial Neural Networks, Barcelona, Spain, September 6-9, 2016, Proceedings, Part II 25. Springer International Publishing, pp. 63-71, 2016.
  13. SIMONYAN, Karen; ZISSERMAN, Andrew, "Very Deep Convolutional Networks for Large-Scale Image Recognition," arXiv preprint arXiv:1409.1556, 2014.
  14. TAN, Mingxing; LE, Quoc, "Efficientnet: Rethinking Model Scaling for Convolutional Neural Networks," In: International Conference on Machine Learning, PMLR pp. 6105-6114, 2019.