论文信息 - Improved product identification method using CNN for mixed-reality web shopping system

Improved product identification method using CNN for mixed-reality web shopping system

E-commerce (EC) sites have become quite popular for online shopping. However, a user of an EC site must possess certain skills to search for his/her desired product from the numerous web pages. A mixed-reality web-shopping system with panoramic views mitigates this problem. The users of this system can shop in the same manner as they do in a real shop, by viewing the panoramic photographs taken inside the shop. However, one problem with this system is the detection of a product selected in the panoramic view by the user. Although we proposed a product identification system using convolutional neural networks (CNN), its performance was not satisfactory. In this research, we propose a method to improve the recognition rate of the CNN.

Kazuki Shibata | Masaya Ohta | Hotaka Niwa

[1] Geoffrey E. Hinton,et al. ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[2] Yoshua Bengio,et al. Gradient-based learning applied to document recognition , 1998, Proc. IEEE.

[3] Qiang Chen,et al. Network In Network , 2013, ICLR.

[4] Katsumi Yamashita,et al. A product identification method for a mixed-reality web shopping system , 2016, 2016 IEEE 5th Global Conference on Consumer Electronics.

[5] Katsumi Yamashita,et al. Mixed-reality web shopping system using panoramic view inside real store , 2015, SIGGRAPH Asia Mobile Graphics and Interactive Applications.

[6] Kota Iwamoto,et al. Local Feature Based Multiple Object Instance Identification Using Scale and Rotation Invariant Implicit Shape Model , 2014, ACCV Workshops.