본문 바로가기
KERD
논문KCI 등재

자율주행 자동차 인지 성능 향상을 위한 복수의 인공신경망 결과 값 합성법

Boosting Transferability of Pre-Trained Models through Multiple Model Fusion

백윤석|이혁기|신성근|신유현
한국ITS학회 (The Korean Society of Intelligent Transport Systems)|2024.12|한국ITS학회 논문지 (Journal of Korean Society of Intelligent Transport Systems)|제23권 제6호|pp. 291-304|DOI https://doi.org/10.12815/kits.2024.23.6.291

0

조회

0

다운로드

0

인용

0

참고문헌

키워드

딥러닝인공신경망Deep learningNeural networkImage classificationTransfer learning전이 학습이미지 인식모델 합성Model fusion

초록

전이 학습은 이미 학습된 딥러닝 모델을 초기 모델로 활용하여, 다른 데이터에서 높은 성능을 발휘하는 기술이다. 특히, 학습에 사용할 데이터의 질과 양이 충분하지 않을 때 전이 학습은매우 유용한 것으로 알려져 있기에 높은 성능 안정성과 많은 데이터를 필요로 하는 자율주행차 인식 분야에 응용될 수 있다. 그러나 많은 선행 연구들은 전이 학습 알고리즘 자체의 성능향상에 초점을 맞추었다. 본 연구는 기존의 알고리즘 중심 접근 방식에서 벗어나, 여러 데이터셋의 특징 추출기 출력을 융합한 심층 모델 융합 기법을 전이 학습에 적용하고자 한다. 실험결과, 이러한 심층 모델 융합 기법이 전이 학습의 성능을 향상시킬 수 있음을 확인했다. 이 결과는 앞으로 데이터가 부족한 자율주행 자동차 분야에서 전이 학습에 활용되어 물체 인지 성능의 향상을 달성할 수 있을 것으로 기대된다.

Abstract

Transfer learning is a technique that leverages a deep learning model trained on a specific dataset as an initial model that allows fast training of a high-performing model on another dataset. Because a pre-trained model already learns how to extract the features from previously trained data, it allows for faster and better performance on new datasets compared to models that are initialized randomly. Transfer learning is particularly useful when the quality and quantity of the data to be learned are insufficient. On the other hand, most studies focused on improving the performance of transfer learning algorithms themselves. This study departs from the existing algorithm-centric research approach and aims to incorporate deep model fusion techniques that combine the outputs of feature extractors from different datasets into transfer learning. These experiments show that the application of deep model fusion improves the performance of transfer learning. These findings will be applicable to transfer learning in various domains with limited data.

참고문헌 (45)

  1. Aniol, A., "Ensemble approach for natural language question answering problem", IEEE 2019 Seventh International Symposium on Computing and Networking Workshops(CANDARW), pp.180~183 (2019)
  2. Borgwardt, K. M., "Integrating structured biological data by kernel maximum mean discrepancy", Bioinformatics, 22(14), pp.49~57 (2006)
  3. Bossard, L., "Food-101 -- Mining discriminative components with random forests", European Conference on Computer Vision (2014)
  4. Breiman, L., "Bagging predictors", Machine Learning, 24, pp.123~140 (1996)
  5. Cimpoi, M., "Describing textures in the wild", Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp.3606~3613 (2014)
  6. Cui, Y., "Dg-labeler and dgl-mots dataset : Boost the autonomous driving perception", Proceedings of the IEEE, 2022, 58~67
  7. Daumé III, H., "Frustratingly easy domain adaptation", Proc. 45th Annual Meeting of the Association for Computational Linguistics, pp.256~263 (2007)
  8. Deng, J., "ImageNet: A large-scale hierarchical image database", 2009 IEEE Conference on Computer Vision and Pattern Recognition, pp.248~255 (2009)
  9. Duan, L., "Learning with augmented features for heterogeneous domain adaptation", Proc. 29th International Conference on Machine Learning, pp.1~8 (2012)
  10. Fort, S., "Large scale structure of neural network loss landscapes", Advances in Neural Information Processing Systems, 32 (2019)
  11. He, K., "Deep residual learning for image recognition", Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp.770~778 (2016)
  12. Hoang, M., "Collective model fusion for multiple black-box experts", International Conference on Machine Learning, pp.2742~2750 (2019)
  13. Hou, C. A., "Unsupervised domain adaptation with label and structural consistency", IEEE Trans. Image Process, 25(12), pp.5552~5562 (2016)
  14. Jung, Y. S., "Class 1·3 vehicle classification using deep learning and thermal image", Journal of the Korea Institute of Intelligent Transportation Systems, 19(6), pp.96~106 (2020)
  15. Kontschieder, P., "Deep neural decision forests", Proceedings of the IEEE International Conference on Computer Vision, pp.1467~1475 (2015)
  16. Krizhevsky, A., "Learning Multiple Layers of Features from Tiny Images" (2009)
  17. Kumar, A., "Co-regularization based semi-supervised domain adaptation", Proc. 24th Annual Conference on Neural Information Processing Systems, pp.478~486 (2010)
  18. Lee, C. Y., "Sliced Wasserstein discrepancy for unsupervised domain adaptation", Proc. IEEE Conference on Computer Vision and Pattern Recognition, pp.10285~10295 (2019)
  19. Leontev, M. I., "Non-iterative knowledge fusion in deepconvolutional neural networks", Neural Processing Letters, 51, pp.1~22 (2020)
  20. Li, W., "Learning with augmented features for supervised and semi-supervised heterogeneous domain adaptation", IEEE Trans. Pattern Anal, 36(6), pp.1134~1148 (2014)
  22. Li, Y., "Ranking neural checkpoints", Proceedings of the IEEE, 2021, 2663~2673
  23. Long, M., "Transfer feature learning with joint distribution adaptation", Proc. IEEE International Conference on Computer Vision, pp.2200~2207 (2013)
  25. Netzer, Y., "Reading digits in natural images with unsupervised feature learning", NIPS Workshop on Deep Learning and Unsupervised Feature Learning, 2011(2), pp.4~ (2011)
  26. Nguyen, C., "Leep: A new measure to evaluate transferability of learned representations", Proceedings of the 37th International Conference on Machine Learning (2020)
  27. Nilsback, M. E., "Automated flower classification over a large number of classes", IEEE 2008 Sixth Indian Conference on Computer Vision, Graphics & Image Processing, pp.722~729 (2008)
  28. Pan, S. J., "Transfer learning via dimensionality reduction", Proc. 23rd AAAI Conference on Artificial Intelligence, pp.677~682 (2008)
  29. Pan, S. J., "Domain adaptation via transfer component analysis", Proc. 21st International Joint Conference on Artificial Intelligence, pp.1187~1192 (2009)
  30. Pan, S. J., "Domain adaptation via transfer component analysis", IEEE Transactions on Neural Network, 22(2), pp.199~210 (2011)
  31. Parkhi, O. M., "Cats and dogs", IEEE Conference on Computer Vision and Pattern Recognition, pp.3498~3505 (2012)
  32. Schapire, R. E., "A brief introduction to boosting", Proceedings of the Sixteenth International Joint Conference on Artificial Intelligence, 99, pp.1401~1406 (1999)
  33. Seo, J., "Application of deep learning-based object detection and distance estimation algorithms for driving to urban area", The Journal of The Korea Institute of Intelligent Transport Systems, 21(3), pp.83~95 (2022)
  34. Shen, J., "Wasserstein distance guided representation learning for domain adaptation", Proc. 32nd AAAI Conference on Artificial Intelligence, pp.4058~4065 (2018)
  35. Singh, S. P., "Model fusion via optimal transport", Advances in Neural Information Processing Systems, 33, pp.22045~22055 (2020)
  36. Stallkamp, J., "Man vs. computer: Benchmarking machine learning algorithms for traffic sign recognition", Neural Networks, 32, pp.323~332 (2012)
  37. Tahmoresnezhad, J., "Visual domain adaptation via transfer feature learning", Knowledge and Information Systems, 50(2), pp.585~605 (2017)
  38. Thomee, B., "Yfcc100m : The new data in multimedia research", Communications of the ACM, 59(2), pp.64~72 (2016)
  39. Wang, B., "A chinese question answering approach integrating count-based and embedding-based features", Natural Language Understanding and Intelligent Applications: 5th CCF Conference on Natural Language Processing and Chinese Computing, pp.934~941 (2016)
  40. Wang, Z., "Meta-learning without data via wasserstein distributionally-robust model fusion", Uncertainty in Artificial Intelligence, pp.2045~2055 (2022)
  41. Weinberger, K. Q., "Learning a kernel matrix for nonlinear dimensionality reduction", Proc. 21st International Conference on Machine Learning, pp.106~113 (2004)
  42. Wolpert, D. H., "Stacked generalization", Neural Networks, 5(2), pp.241~259 (1992)
  43. Zellinger, W., "Central moment discrepancy (CMD) for domaininvariant representation learning", Proc. 5th International Conference on Learning Representations, pp.1~13 (2017)
  44. Zhang, J., "Joint geometrical and statistical alignment for visual domain adaptation", Proc. IEEE Conference on Computer Vision and Pattern Recognition, pp.5150~5158 (2017)
  45. Zhou, Z. H., "Deep forest", National Science Review, 6(1), pp.74~86 (2019)

관련 논문AI추천

자율주행을 위한 인공지능 기술 동향 및 발전 방향김병도, 김재겸, 백재순, 유진혁, 최준원한국모빌리티학회 · 모빌리티연구2021 · 제1권 제2호자율주행 자동차의 주행을 위한 딥러닝 모델 연구문병준, 하성재한국산업기술융합학회 · 산업기술연구논문지2024 · 제29권 제3호블록 계층별 재학습을 이용한 다중 힌트정보 기반 지식전이 학습배지훈대한임베디드공학회 · 대한임베디드공학회논문지2020 · 제15권 제2호전이 학습을 이용한 자동차 페달 블랙박스 영상 분류 연구임종범, 최희석한국정보처리학회 · 정보처리학회 논문지2025 · 제14권 제2호자율주행 차량의 학습 데이터 자동 생성 시스템 개발홍준, 김재환, 김형주, 윤승제, 임경일, 정지원한국ITS학회 · 한국ITS학회 논문지2020 · 제19권 제5호국내 도로 환경에 특화된 자율주행을 위한 멀티카메라 데이터 셋 구축 및 유효성 검증권순, 김제석, 박재형, 이재근, 이진희대한임베디드공학회 · 대한임베디드공학회논문지2022 · 제17권 제5호

이 간행물의 다른 논문 (25건)

도로환경 교통사고 상해도 예측을 위한 iGLAD 기반 머신러닝 모델 성능 비교김형규, 최용순, 김천호, 정승윤, 백세룡pp. 1-12빅 데이터를 이용한 승용-버스 간의 이동성 특성 분석: 5개 광역시를 중심으로이규진, 정우현, 이승찬, 류인곤pp. 13-27고속도로 2차 사고 심각도 영향요인 분석이재덕, 장일준, 안세영pp. 28-42XAI 기법을 활용한 트립 체인 데이터 기반 화물운전자 통행 식별 모델 개발이광섭, 이상수, 이요셉, 김인영, 윤일수pp. 43-62실사고 데이터 기반 자율주행자동차 교통사고 심각도 영향요인 분석박기옥, 이현미, 윤일수, 이요셉, 장정아pp. 63-80
원문 보기