Learning AND-OR Templates for Professional Photograph Parsing and Guidance

TL;DR

Experimental results show that the learned templates can well describe the photography techniques and styles, whereas the proposed approach can assess the quality of photography images as human being does.

Abstract

Since the development of photography art, many so-called "templates" have been formed, namely visual styles summarized from a series of themed and stylized photography works. In this paper, we propose to analysize and and summarize these 'templates' in photography by learning composite templates of photography images. We present a framework for learning a hierarchical reconfigurable image template from photography images to learn and characterize the "templates" used in these photography images. Using this method, we measured the artistic quality of photography on the photos and conducted photography guidance. In addition, we also utilized the "templates" for guidance in several image generation tasks. Experimental results show that the learned templates can well describe the photography techniques and styles, whereas the proposed approach can assess the quality of photography images as human being does.

Figures (6)

• Figure 1: The process of summarizing the 'template' by photographers. Photographers can summarize templates from a series of themed and stylized photography works, and then use these templates as a reference to improve and innovate, forming their own unique shooting style and aesthetic.
• Figure 2: The detailed illustration of our composite template. For OR nodes, we illustrate typical templates with structural variations and geometric transforms. The AND nodes are denoted by solid blue circles. The terminal nodes are individual parts which are represented by professional photography templates.
• Figure 3: Overall process of template learning.
• Figure 4: Initial Data Matrix. Data matrix R is a matrix with N (number of positive examples) rows and D (number of all candidate features) columns, and each entry $R_{ij}$ = $r_j$($I_i$) is a feature response (0 $\leq$$R_{ij}$$\leq$ 1). Larger value of $R_{ij}$ means feature j appears in image $I_i$ with higher probability.
• Figure 5: Results of quantitative assessment. In addition to being able to output the matching score between the image and the template, we can also output the configuration of the image on the template.