Journal of the Institute of Convergence Signal Processing (융합신호처리학회논문지)

The Korea Institute of Convergence Signal Processing (한국융합신호처리학회)
권호 표지
DOI QR코드

DOI QR Code

Classification of Raccoon dog and Raccoon with Transfer Learning and Data Augmentation

전이 학습과 데이터 증강을 이용한 너구리와 라쿤 분류

  • Dong-Min Park (Department of IT Convergence Engineering, Daegu University) ;
  • Yeong-Seok Jo (Department of Biology Education, Daegu University) ;
  • Seokwon Yeom (Department of IT Convergence Engineering, Daegu University)
  • 박동민 (대구대학교 IT융합공학과) ;
  • 조영석 (대구대학교 생물교육학과) ;
  • 염석원 (대구대학교 IT융합공학과)
  • Received : 2022.11.24
  • Accepted : 2023.03.05
  • Published : 2023.03.31
Download PDF
⟨ Previous Next ⟩

Abstract

In recent years, as the range of human activities has increased, the introduction of alien species has become frequent. Among them, raccoons have been designated as harmful animals since 2020. Raccoons are similar in size and shape to raccoon dogs, so they generally need to be distinguished in capturing them. To solve this problem, we use VGG19, ResNet152V2, InceptionV3, InceptionResNet and NASNet, which are CNN deep learning models specialized for image classification. The parameters to be used for learning are pre-trained with a large amount of data, ImageNet. In order to classify the raccoon and raccoon dog datasets as outward features of animals, the image was converted to grayscale and brightness was normalized. Augmentation methods were applied using left and right inversion, rotation, scaling, and shift to create sufficient data for transfer learning. The FCL consists of 1 layer for the non-augmented dataset while 4 layers for the augmented dataset. Comparing the accuracy of various augmented datasets, the performance increased as more augmentation methods were applied.

최근 인간의 활동 범위가 증가함에 따라 외래종의 유입이 잦아지고 있고 환경에 적응하지 못해 유기된 외래종 중 2020년부터 유해 지정 동물로 지정된 라쿤이 문제가 되고 있다. 라쿤은 국내 토종 너구리와 크기나 생김새가 유사하여 일반적으로 포획하는데 있어서 구분이 필요하다. 이를 해결하기 위해서 이미지 분류에 특화된 CNN 딥러닝 모델인 VGG19, ResNet152V2, InceptionV3, InceptionResNet, NASNet을 사용한다. 학습에 사용할 파라미터는 많은 양의 데이터인 ImageNet으로 미리 학습된 파라미터를 전이 학습하여 이용한다. 너구리와 라쿤 데이터셋에서 동물의 외형적인 특징으로 분류하기 위해서 이미지를 회색조로 변환한 후 밝기를 정규화하였으며, 조정된 데이터셋에 충분한 학습을 위한 데이터를 만들기 위해 좌우 반전, 회전, 확대/축소, 이동을 이용하여 증강 기법을 적용하였다. 증강하지 않은 데이터셋은 FCL을 1층으로, 증강된 데이터셋은 4층으로 구성하여 진행하였다. 여러 가지 증강된 데이터셋의 정확도를 비교한 결과, 증강을 많이 할수록 성능이 증가함을 확인하였다.

Keywords

Acknowledgement

본 결과물은 환경부의 재원으로 한국환경산업기술원의 생물다양성위협 외래생물관리기술개발사업의 지원을 받아 연구되었습니다.(KEITI 2021002270003).

References

  1. Kim, D. E, "Management System of Invasive Alien Species Threating Biodiversity in Korea and Suggestions for the Improvement", Journal of Environmental Impact Assessment, 27(1), 33-55(2018). https://doi.org/10.14249/EIA.2018.27.1.33
  2. Ministry of Environment. 2014. 1st Management Plan of Alien Species (2014-2018).
  3. Da Bin Kim. Kyung Ah Koo. "A Study on the Current Status and Improvement of Ecosystem Disturbance Species", 29(4), 59-81(2021). https://doi.org/10.15301/jepa.2021.29.4.59
  4. Ikeda T, Asano M, Matoba Y, Abe G, "Present status of invasive alien raccoon and its impact in Japan", Global environmental research, 8(2), pp. 125-131, 2004.
  5. National Institute of Ecology, "Ecological studies of alien species (VI)", Seocheon National Institute of Ecology
  6. Jihae Kim et al, "Comparison of Fine Grained Classification of Pet Images Using Image Processing and CNN", Journal of Broadcast Engineering (JBE) Vol. 26, No. 2, pp. 175-183, 2021.
  7. Barret Zoph, Vijay Vasudevan, Jonathon Shlens, Quoc V. Le; "Learning Transferable Architectures for Scalable Image Recognition", Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 8697-8710
  8. Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun; "Deep Residual Learning for Image Recognition", Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 770-778
  9. Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jon Shlens, Zbigniew Wojna; "Rethinking the Inception Architecture for Computer Vision", Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 2818-2826
  10. Christian Szegedy, Sergey Ioffe, Vincent Vanhoucke, Alexander A Alemi; "Inception-v4, inception-resnet and the impact of residual connections on learning", Thirty-first AAAI conference on artificial intelligence. 2017.
  11. Simonyan, Karen, and Andrew Zisserman; "Very deep convolutional networks for large-scale image recognition", arXiv preprint arXiv:1409.1556 (2014).
  12. Kaiming He Ross Girshick Piotr Doll'ar, "Rethinking ImageNet Pre-training" arXiv:1811.08883v1 (2018).
  13. Shorten, C., Khoshgoftaar, T.M., "A survey on Image Data Augmentation for Deep Learning", J Big Data 6, 60 (2019). https://doi.org/10.1186/s40537-019-0197-0
  14. J. Deng, W. Dong, R. Socher, L. -J. Li, Kai Li and Li Fei-Fei, "ImageNet: A large-scale hierarchical image database", 2009 IEEE Conference on Computer Vision and Pattern Recognition, Miami, FL, USA, 2009, pp. 248-255, doi: 10.1109/CVPR.2009.5206848
  15. "Keras Applications,"Keras, https://keras.io/api/applications/