대형 이미지 데이터셋 구축을 위한 객체 엣지 기반이미지 생성 기법
Object Edge-based Image Generation Technique for Constructing Large-scale Image Datasets
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
딥러닝의 발전은 컴퓨터 비전 문제를 해결할 수 있지만, 높은 정확도를 위해서는 대규모 데이터셋이 필요하다. 본 논문에서는 객체 바운딩 박스와 이미지 엣지 성분을 이용한 이미지 생성 기법을 제안한다. 객체 탐지를 통해 이미지 내의 객체 바운딩 박스를 추출하고 이미지 엣지 성분을 함께 이미지 생성모델의 입력값으로 사용하여 새로운 이미지 데이터를 생성한다. 실험 결과, 제안 기법으로 생성된 이미지는 이미지 품질 평가에서 소스 이미지와 유사한 품질을 보였고, 딥러닝 훈련과정에서도 좋은 성능을 보였다
Abstract
Deep learning advancements can solve computer vision problems, but large-scale datasets are necessary for highaccuracy. In this paper, we propose an image generation technique using object bounding boxes and image edgecomponents. The object bounding boxes are extracted from the images through object detection, and image edgecomponents are used as input values for the image generation model to create new image data. As results ofexperiments, the images generated by the proposed method demonstrated similar image quality to the sourceimages in the image quality assessment, and also exhibited good performance during the deep learning trainingprocess.
참고문헌 (15)
- LeCun Y, "Deep learning", Nature, 521, pp.436~444 (2015)
- Connor Shorten, "A survey on Image Data Augmentation for Deep Learning", Journal of Big Data, 6(1) (2019), Springer Science and Business Media LLC[DOI]
- C. Sun, "Revisiting Unreasonable Effectiveness of Data in Deep Learning Era", 2017 IEEE International Conference on Computer Vision (ICCV), pp.843~852 (2017)
- Athanasios Voulodimos, "Deep Learning for Computer Vision: A Brief Review", Computational Intelligence and Neuroscience, pp.13~ (2018)
- T. Karras, "A Style-Based Generator Architecture for Generative Adversarial Networks", 2019 IEEE, 2019, 4396~4405
- Ju-hyeok Lee, "Synthetic data generation technique using object bounding box and original image combination", Conference of korea infomation processing society, 30(1), pp.476~478 (2023)
- Perez, Luis, "The Effectiveness of Data Augmentation in Image Classification using Deep Learning", Computer Vision and Pattern Recognition, pp.8~ (2017)[DOI]
- Alex Krizhevsky, "ImageNet classification with deep convolutional neural networks", Advances in Neural Information Processing Systems (NIPS), pp.1097~1105 (2012)
- N. V. Chawla, "SMOTE: Synthetic Minority Over-Sampling Technique", Journal of Artificial Intelligence Research, 16, pp.321~357 (2002)[DOI]
- Diederik P. Kingma, "Auto-Encoding variational bayes", International conference on Learning `Representations (ICLR), pp.14~ (2014)
- Goodfellow, I., "Generative adversarial nets", Advances in Neural Information Processing Systems (NIPS), pp.2672~2680 (2014)
- Gao, Wenshuo, "An improved Sobel edge detection", 2010 3rd International conference on computer science and information technology, 5 (2010)
- "kaggle, Cats and Dogs dataset to train a DL model"[URL]
- Yuanji. W, "Image quality evaluation based on image weighted separating block peak signal to noise ratio", International Conference on Neural Networks and Signal Processing, 2, pp.994~997 (2003), IEEE
- Z. Wang, "Image Quality Assessment: From Error Visibility to Structural Similarity", IEEE Transactions on Image Processing, 13(4), pp.600~612 (2004), Institute of Electrical and Electronics Engineers (IEEE)[DOI]