Title of Presentation

“Visualizing Direction of Attention to Understand AI Decisions”

Artificial intelligence systems using deep learning have achieved image and speech recognition at performance levels on par with human beings. However, an issue is not being able to understand the basis on which a deep learning system determines its output. In this lecture, we introduce Attention Branch Network (ABN), which outputs attention, an area that deep learning focuses on when determining inferential results. ABN is a deep learning network that can contribute to improving recognition performance while acquiring the attention mechanism. As application examples of ABN, we introduce ABN’s visual explanation of automated driving and medical diagnosis decisions. Visualization of attention here means being able to view the AI’s direction of attention. This technology holds great promise as an approach for interpreting the basis for decisions output by an AI system.

Profile

Web Site URL

http://mprg.jp/

A brief Biography(As of April 1, 2019)

1997	Ph.D., Chubu University
1997	Postdoctoral Fellow, Robotics Institute, Carnegie Mellon University
2000	Senior Assistant Professor, Department of Computer Science, College of Engineering, Chubu University
2004	Associate Professor, Chubu University
2005 – 2006	Visiting Researcher, Robotics Institute, Carnegie Mellon University
2010 – Present	Professor, Chubu University
2014 – Present	Visiting Professor, Nagoya University

Details of selected Awards and Honors

2005	RoboCup Research Prize
2009	Information Processing Society of Japan Transactions on Computer Vision and Image Media Outstanding Paper Award
2009	Information Processing Society of Japan Yamashita SIG Research Award
2010	Symposium on Sensing via Image Information Outstanding Academic Award
2013	Institute of Electronics, Information and Communication Engineers Information and Systems Society Excellent Paper Award

A list of selected Publications

H. Fukui, T. Hirakawa, T. Yamashita and H. Fujiyoshi, “Attention Branch Network: Learning of Attention Mechanism for Visual Explanation”, IEEE/CVF International Conference on Computer Vision and Pattern Recognition, 2018.

T. Hasegawa, M. Ambai, K. Ishikawa, G. Koutaki, Y. Yamauchi, T. Yamashita and H. Fujiyoshi, “Multiple-hypothesis affine region estimation with anisotropic LoG filters”, International Conference on Computer Vision, pp. 585-593, 2015.

Goto, Y., Tsuchiya, M., Yamauchi, Y., and Fujiyoshi, H. Fast discrimination by early judgment using linear classifier based on approximation calculation, IEICE Transactions [on Information and Systems], Vol. J97-D, No.2, pp.294-302, 2014.

Mitsui, T., Yamauchi, Y., and Fujiyoshi, H. Object detection by two-stage boosting with joint features, IEICE Transactions [on Information and Systems], Vol. J92-D, no. 9, pp. 1591–1601, 2009.

Nagahashi, T., Fujiyoshi, H., and Kanade, T. Image segmentation using iterated graph cuts based on multi-scale smoothing, IPSJ Transactions on Computer Vision and Image Media Vol.1, No.2, pp.10-20, 2008.

Y. Yamauchi, H. Fujiyoshi, Y. Iwahori, and T. Kanade, “People Detection Based on Co-occurrence of Appearance and Spatio-temporal Features”, National Institute of Informatics Transactions on Progress in Informatics, no. 7, pp. 33–42, 2010.

H. Fujiyoshi and T. Kanade, “Layered Detection for Multiple Overlapping Objects”, IEICE TRANSAC- TIONS on Information and Systems, vol. E87-D, pp. 2821–2827, 2004.

H. Fujiyoshi, A. J. Lipton, and T. Kanade, “Real-Time Human Motion Analysis by Image Skeletonization”, IEICE TRANSACTIONS on Information and Systems, vol. E87-D, pp. 113–120, 2004.

R. T. Collins, A. J. Lipton, H. Fujiyoshi, and T. Kanade, “Algorithms for cooperative multisensor surveillance”, IEEE vol. 89, pp. 1456–1477, 2001.