[3D5-OS-22b] OS-22 (2)

[3D5-OS-22b-02] Photograph creation analysis based multiple evaluation features to acquire filming techniques

〇Toshinori Aoki¹, Miki Ueno¹ (1. Osaka Institute of Technology )

Keywords:Photograph creation, Filmcraft support, Machine Learning

Recently, numerous people take photographs easily by smart-phones and post them on social network services. On the other hand, huge knowledge and technique are required in order to become skilled photographers. It is hard to acquire these information especially on real-time filming.
Thus, the aim of the research is to create a system to suggest advices for novices of photographers.To construct such a system, the method to evaluate photos precisely like a professional photographer is required.
In this paper, we have proposed original dataset of scenery photographs with detailed scores by professional photographers and construct Deep Convolutional Neural Networks to classify two class images with the labels evaluated by the author as a professional photographer. In addition, we consider acquired features by Gradient Weighted Class Activation Mapping.