Computer Vision Research and Development in Sony

Computer vision technologies have been implemented on many Sony products and brought a huge amount of value for over 20 years. I will introduce image recognition applications in Sony products and recent research beyond visual-only recognition. Image recognition for consumer electronics should run on the edge for the users’ privacy, fewer delays, and power efficiency. This talk will give discussions based on the actual cases. Multimodal recognition is also highly demanded in these application fields beyond visual-only recognition systems. We have contributed to audio-visual scene recognition by proposing quantitative metrics and designing baseline models for it. I will show how our models perform well in challenging situations such as multiple-sounding objects.

Daiki Shimada is a research engineer at R&D Center of Sony Group Corporation. He is now visiting the CS department at University of Rochester as a visiting researcher. His research interests are machine perception for robots and multimodal scene understanding. He has been involved in various products/projects that have used computer vision technology in Sony, including Spatial Reality Display and Xperia smartphone. He received BS and MS degrees in applied informatics from Hosei University, Japan, in 2015 and 2017. His honors include Sony R&D Center Award 2021 bronze, The Japanese Society for Artificial Intelligence 30th Anniversary Project Award, and Honorable Award YANS Symposium 2016.

Lunch will be served.