Skip to main content
main-content
Top

Hint

Swipe to navigate through the chapters of this book

2021 | OriginalPaper | Chapter

Development of Fast Refinement Detectors on AI Edge Platforms

share
SHARE

Abstract

Refinement detector (RefineDet) is a state-of-the-art model in object detection that has been developed and refined based on high-end GPU systems. In this study, we discovered that the speed of models developed in high-end GPU systems is inconsistent with that in embedded systems. In other words, the fastest model that operates on high-end GPU systems may not be the fastest model on embedded boards. To determine the reason for this phenomenon, we performed several experiments on RefineDet using various backbone architectures on three different platforms: NVIDIA Titan XP GPU system, Drive PX2 board, and Jetson Xavier board. Finally, we achieved real-time performances (approximately 20 fps) based on the experiments on AI edge platforms such as NVIDIA Drive PX2 and Jetson Xavier boards. We believe that our current study would serve as a good reference for developers who wish to apply object detection algorithms to AI edge computing hardware. The complete code and models are publicly available on the web (link).
Literature
2.
go back to reference Chollet, F.: Xception: deep learning with depthwise separable convolutions. In: Proceedings of CVPR (2017) Chollet, F.: Xception: deep learning with depthwise separable convolutions. In: Proceedings of CVPR (2017)
3.
go back to reference He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: Proceedings of CVPR (2016) He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: Proceedings of CVPR (2016)
4.
go back to reference Hu, J., Shen, L., Sun, G.: Squeeze-and-excitation networks. In: Proceedings of CVPR (2018) Hu, J., Shen, L., Sun, G.: Squeeze-and-excitation networks. In: Proceedings of CVPR (2018)
5.
go back to reference Jin, A., et al.: Tool detection and operative skill assessment in surgical videos using region-based convolutional neural networks. In: Proceedings of WACV (2018) Jin, A., et al.: Tool detection and operative skill assessment in surgical videos using region-based convolutional neural networks. In: Proceedings of WACV (2018)
6.
go back to reference Jung, H., Choi, M.K., Jung, J., Lee, J.H., Kwon, S., Jung, W.Y.: ResNet-based vehicle classification and localization in traffic surveillance systems. In: Proceedings of CVPRW (2017) Jung, H., Choi, M.K., Jung, J., Lee, J.H., Kwon, S., Jung, W.Y.: ResNet-based vehicle classification and localization in traffic surveillance systems. In: Proceedings of CVPRW (2017)
7.
go back to reference Lin, T.Y., et al.: Microsoft COCO: common objects in context. In: Proceedings of ECCV (2014) Lin, T.Y., et al.: Microsoft COCO: common objects in context. In: Proceedings of ECCV (2014)
8.
go back to reference Liu, W., et al.: SSD: single shot multibox detector. In: Proceedings of ECCV (2017) Liu, W., et al.: SSD: single shot multibox detector. In: Proceedings of ECCV (2017)
9.
go back to reference Liu, L., Ouyang, W., Wang, X., Fieguth, P., Liu, X., Pietikainen, M.: Deep learning for generic object detection: a survey. arxiv:​1809.​02165 (2018) Liu, L., Ouyang, W., Wang, X., Fieguth, P., Liu, X., Pietikainen, M.: Deep learning for generic object detection: a survey. arxiv:​1809.​02165 (2018)
10.
go back to reference Liu, W., Rabinovich, A., Berg, A.C.: ParseNet: looking wider to see better. In: Proceedings of ICLR (2016) Liu, W., Rabinovich, A., Berg, A.C.: ParseNet: looking wider to see better. In: Proceedings of ICLR (2016)
11.
go back to reference Russakovsky, O., et al.: ImageNet large scale visual recognition challenge. Int. J. Comput. Vis. 115(3), 211–252 (2015) MathSciNetCrossRef Russakovsky, O., et al.: ImageNet large scale visual recognition challenge. Int. J. Comput. Vis. 115(3), 211–252 (2015) MathSciNetCrossRef
12.
go back to reference Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., Chen, L.-C.: MobileNetV2: inverted residuals and linear bottlenecks. In: Proceedings of CVPR (2018) Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., Chen, L.-C.: MobileNetV2: inverted residuals and linear bottlenecks. In: Proceedings of CVPR (2018)
13.
go back to reference Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., Wojna, Z.: Rethinking the inception architecture for computer vision. In: Proceedings of CVPR (2016) Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., Wojna, Z.: Rethinking the inception architecture for computer vision. In: Proceedings of CVPR (2016)
14.
go back to reference Wu, B., Wan, A., Iandola, F., Jin, P.H., Keutzer, K.: SqueezeDet: unified, small, low power fully convolutional neural networks for real-time object detection for autonomous driving. In: Proceedings of CVPR (2017) Wu, B., Wan, A., Iandola, F., Jin, P.H., Keutzer, K.: SqueezeDet: unified, small, low power fully convolutional neural networks for real-time object detection for autonomous driving. In: Proceedings of CVPR (2017)
15.
go back to reference Xie, S., Girshick, R., Dollar, P., Tu, Z., He, K.: Aggregated residual transformations for deep neural networks. In: Proceedings of CVPR (2017) Xie, S., Girshick, R., Dollar, P., Tu, Z., He, K.: Aggregated residual transformations for deep neural networks. In: Proceedings of CVPR (2017)
16.
go back to reference Zhang, S., Wen, L., Bian, X., Lei, Z., Li, S.Z.: Single-shot refinement neural network for object detection. In: Proceedings of CVPR (2018) Zhang, S., Wen, L., Bian, X., Lei, Z., Li, S.Z.: Single-shot refinement neural network for object detection. In: Proceedings of CVPR (2018)
17.
go back to reference Yao, Y., et al.: End-to-end convolutional neural network model for gear fault diagnosis based on sound signals. Appl. Sci. 8(9), 1584 (2018) CrossRef Yao, Y., et al.: End-to-end convolutional neural network model for gear fault diagnosis based on sound signals. Appl. Sci. 8(9), 1584 (2018) CrossRef
18.
go back to reference Redmon, J., Farhadi, A.: YOLO 9000: better, faster, stronger. In: Proceedings of CVPR (2017) Redmon, J., Farhadi, A.: YOLO 9000: better, faster, stronger. In: Proceedings of CVPR (2017)
19.
go back to reference Girshick, R., Donahue, J., Darrell, T., Malik, J.: Rich feature hierarchies for accurate object detection and semantic segmentation. In Proceedings of CVPR (2014) Girshick, R., Donahue, J., Darrell, T., Malik, J.: Rich feature hierarchies for accurate object detection and semantic segmentation. In Proceedings of CVPR (2014)
20.
go back to reference Girshick, R.: Fast R-CNN: towards real-time object detection with region proposal networks. In: Proceedings of ICCV (2015) Girshick, R.: Fast R-CNN: towards real-time object detection with region proposal networks. In: Proceedings of ICCV (2015)
21.
go back to reference Ren, S., He, K., Girshick, R., Sun, J.: Faster R-CNN: towards real-time object detection with region proposal networks. In: Proceedings of NIPS (2015) Ren, S., He, K., Girshick, R., Sun, J.: Faster R-CNN: towards real-time object detection with region proposal networks. In: Proceedings of NIPS (2015)
22.
go back to reference Dai, J., Li, Y., He, K., Sun, J.: R-FCN: object detection via region-based fully convolutional networks. In: Proceedings of NIPS (2016) Dai, J., Li, Y., He, K., Sun, J.: R-FCN: object detection via region-based fully convolutional networks. In: Proceedings of NIPS (2016)
23.
go back to reference Dai, J., et al.: Deformable convolutional networks. In: Proceedings of ICCV (2017) Dai, J., et al.: Deformable convolutional networks. In: Proceedings of ICCV (2017)
24.
go back to reference He, K., Gkioxari, G., Dollar, P., Girshick, R.: Mask R-CNN. In: Proceedings of ICCV (2017) He, K., Gkioxari, G., Dollar, P., Girshick, R.: Mask R-CNN. In: Proceedings of ICCV (2017)
26.
go back to reference Lin, T.Y., Goyal, P., Girshick, R., He, K., Dallor, P.: Focal loss for dense object detection. In: Proceedings of ICCV (2017) Lin, T.Y., Goyal, P., Girshick, R., He, K., Dallor, P.: Focal loss for dense object detection. In: Proceedings of ICCV (2017)
27.
go back to reference Shrivastava, A., Gupta, A., Girshick, R.: Training region-based object detectors with online hard example mining. In: Proceedings of CVPR (2016) Shrivastava, A., Gupta, A., Girshick, R.: Training region-based object detectors with online hard example mining. In: Proceedings of CVPR (2016)
28.
go back to reference Bell, S., Zitnick, C.L., Bala, K., Girshick, R.: Inside-outside net: detecting objects in context with skip pooling and recurrent neural networks. In: Proceedings of CVPR (2016) Bell, S., Zitnick, C.L., Bala, K., Girshick, R.: Inside-outside net: detecting objects in context with skip pooling and recurrent neural networks. In: Proceedings of CVPR (2016)
29.
go back to reference Zhu, Y., Zhao, C., Wang, J., Zhao, X., Wu, Y., Lu, H.: CoupleNet: coupling global structure with local parts for object detection. In: Proceedings of ICCV (2017) Zhu, Y., Zhao, C., Wang, J., Zhao, X., Wu, Y., Lu, H.: CoupleNet: coupling global structure with local parts for object detection. In: Proceedings of ICCV (2017)
30.
go back to reference Bodla, N., Singh, B., Chellappa, R., Davis, L.S.: Improving object detection with one line of code. In: Proceedings of ICCV (2017) Bodla, N., Singh, B., Chellappa, R., Davis, L.S.: Improving object detection with one line of code. In: Proceedings of ICCV (2017)
31.
go back to reference Huang, J., et al.: Speed/accuracy tradeoffs for modern convolutional object detectors. In: Proceedings of CVPR (2017) Huang, J., et al.: Speed/accuracy tradeoffs for modern convolutional object detectors. In: Proceedings of CVPR (2017)
32.
go back to reference Kong, T., Sun, F., Yao, A., Liu, H., Lu, M., Chen, Y.: RON reverse connection with objectness prior networks for object detection. In: Proceedings of CVPR (2017) Kong, T., Sun, F., Yao, A., Liu, H., Lu, M., Chen, Y.: RON reverse connection with objectness prior networks for object detection. In: Proceedings of CVPR (2017)
Metadata
Title
Development of Fast Refinement Detectors on AI Edge Platforms
Authors
Min-Kook Choi
Heechul Jung
Copyright Year
2021
DOI
https://doi.org/10.1007/978-3-030-68799-1_43

Premium Partner