Abstract

Object detection and data collection from custom targets suffer from certain problems, which inherently occur in deep learning networks owing to problems such as difficulty of collection and data bias. Therefore, in this study, we proposed the Multi-GAN framework for generating augmented datasets. This framework comprises two parts: the first part generates data that reflect various textures related to decep learning based on deep convolutional GAN (DCGAN) and Wasserstein GAN (WGAN) structures. The second part provides multiple resolutions based on super-resolution GAN (SRGAN). Here, this paper presents efficient dataset construction methods along with a conventional augmentation method called manipulation technique. Through the experiments, which were based on average precision, conducted on the collected and augmented datasets, the proposed frameworkdemonstrated to improve detection accuracy. Additionally, we confirmed that the multi-GAN framework is superior with respect to efficiency to data collection.

References

• Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E. Hinton, "Imagenet classification with deep convolutional neural networks," in Proc. NIPS, 2012, pp. 1097-1105,
• Redmon, Joseph, Santosh Divvala, Ross Girshick and Ali Farhadi, “You only look once: Unified, real-time object detection,” in Proc. CVPR, 2016, pp. 779-788, 10.1109/CVPR.2016.91.
• Wei Liu et al, Dragomir Anguelov, Dumitru Erhan, Christian Szegedy, Scott Reed, Cheng-Yang Fu and Alexander C. Berg, “Ssd: Single shot multibox detector,” in European conference on computer vision. Springer, Cham, 2016, pp. 21-37, 10.1007/978-3-319-46448-0_2.
• R. Girshick et al, “Rich feature hierarchies for accurate object detection and semantic segmentation,” in Proc. CVPR, 2014, pp. 580-587, 10.1109/CVPR.2014.81.
• Terrance DeVries and Graham W Taylor, “Improved regularization of convolutional neural networks with cutout,” 2017, arXiv preprint arXiv:1708.04552.
• Sangdoo Yun, Dongyoon Han, Seong Joon Oh, Sanghyuk Cheo and Youngjoon Yoo, “CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Features,” in Proc. ICCV, 2019, pp. 6023-6032, 10.1109/ICCV.2019.00612.
• Wang Hao and Song Zhili, “Improved Mosaic: Algorithms for more Complex Images,” In: Journal of Physics: Conference Series. IOP Publishing, 2020, pp. 012094, 10.1088/1742-6596/1684/1/012094.
• Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He and Piotr Dollár, “Focal loss for dense object detection,” in Proc. ICCV, 2017, pp. 2980-2988, 10.1109/ICCV.2017.324.
• Mingxing Tan and Quoc V Le, “EfficientNet: Rethinking model scaling for convolutional neural networks,” in Proc. the IEEE international conference on computer vision, 2017, pp. 2980-2988, arXiv:1905.11946.
• R. Girshick, Jeff Donahue, Trevor Darrell and Jitendra Malik, "Rich feature hierarchies for accurate object detection and semantic segmentation,” in Proc. CVPR, 2014, pp. 580-587, 10.1109/CVPR.2014.81.
• R. Girshick, "Fast R-CNN," in Proc. the IEEE conference on computer vision and pattern recognition, 2015, arXiv:1504.08083.
• Shaoqing Ren, Kaiming He, Ross Girshick and Jian Sun, “Faster R-CNN: Towards real-time object detection with region proposal networks,” Advances in neural information processing systems, 2015, pp. 91-99, 10.1109/TPAMI.2016.2577031.
• Kaiming He, Georgia Gkioxari, Piotr Dollár and Ross Girshick, “Mask R-CNN,” in Proc. ICCV, 2017, pp. 2961-2069, arXiv:1703.06870.
• Saining Xie, Ross Girshick, Piotr Dollár and Kaiming He, “Aggregated residual transformations for deep neural networks,” in Proc. CVPR, 2017, pp. 1492-1500, 10.1109/CVPR.2017.634.
• Joseph Redmon and Ali Farhadi, “YOLOv3: An incremental improvement,” 2018, arXiv preprint arXiv:1804.02767.
• Alexey Bochkovskiy, Chien-Yao Wang and Hong-Yuan Mark Liao, “YOLOv4: Optimal Speed and Accuracy of Object Detection,” 2020, arXiv preprint arXiv:2004.10934.
• Diganta Misra, “Mish: A self regularized nonmonotonic neural activation function,” 2019, arXiv preprint arXiv:1908.08681.
• Golnaz Ghiasi, Tsung-Yi Lin, and Quoc V Le, “DropBlock: A regularization method for convolutional networks,” 2018, arXiv preprint arXiv:1810.12890.
• Chia-Jung Chou, Jui-Ting Chien and Hwann-Tzong Chen, “Self adversarial Training for Human Pose Estimation,” in 2018 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC). IEEE, 2018, pp. 17-30, arXiv:1707.02439.
• Chien-Yao Wang, Hong-Yuan Mark Liao, Yueh-Hua Wu, Ping-Yang Chen, Jun-Wei Hsieh and I-Hau Yeh, “CSPNet: A new backbone that can enhance learning capability of cnn,” in Proc. CVPR, 2020, pp. 390-391, 10.1109/CVPRW50498.2020.00203.
• Kaiming He, Xiangyu Zhang, Shaoqing Ren and Jian Sun, “Spatial pyramid pooling in deep convolutional networks for visual recognition,” IEEE transactions on pattern analysis and machine intelligence, 2015, 37.9: 1904-1916, 10.1109/TPAMI.2015.2389824.
• Shu Liu, Lu Qi, Haifang Qin, Jianping Shi and Jiaya Jia, “Path aggregation network for instance segmentation,” in Proc. CVPR, 2018, pp. 8759–8768, 10.1109/CVPR.2018.00913.
• Sanghyun Woo, Jongchan Park , Joon-Young Lee and In So Kweon, “CBAM: Convolutional block attention module,” in Proc. ECCV, 2018, pp. 3–19, https://doi.org/10.1007/978-3-030-01234-2_1.
• Zhuliang Yao, Yue Cao, Shuxin Zheng, Gao Huang and Stephen Lin, “Cross-iteration batch normalization,”arXiv preprint arXiv:2002.05712, 2020, 10.1109/CVPR46437.2021.01215.
• Ilya Loshchilov and Frank Hutter, “SGDR: Stochastic gradient descent with warm restarts,” 2016, arXiv preprint arXiv:1608.03983.
• Zhaohui Zheng, Ping Wang, Wei Liu, Jinze Li, Rongguang Ye and Dongwei Ren, “Distance-IoU Loss: Faster and better learning for bounding box regression,” in Proc. the AAAI Conference on Artificial Intelligence, 2020, pp. 12993-13000, arXiv:1911.08287.
• Ian J. Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville and Yoshua Bengio, “Generative adversarial nets,” Advances in neural information processing systems, 2014, 27, https://doi.org/10.1145/3422622.
• Alec Radford, Luke Metz and Soumith Chintala, “Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks,” 2015, arXiv preprint arXiv:1511.06434.
• Martin Arjovsky, Soumith Chintala and Leon Bottou, “Wasserstein GAN,” 2017, arXiv preprint arXiv:1701.07875.
• Jun-Yan Zhu, Taesung Park, Phillip Isola and Alexei A. Efros, “Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks,” 2017, arXiv present  arXiv:1703.10593.
• Christian Ledig, Lucas Theis, Ferenc Huszar, Jose Caballero, Andrew Cunningham, Alejandro Acosta, Andrew Aitken, Alykhan Tejani, Johannes Totz, Zehan Wang and Wenzhe Shi, “Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network,” in Proc. the IEEE conference on computer vision and pattern recognition, 2017, pp. 4681-4690, 10.1109/CVPR.2017.19.