Photo Aesthetics Ranking Network with Attributes and Content Adaptation — Supplementary Material

In this supplementary material, we first present in detail on collecting our AADB dataset, in Section 2, and analyze the dataset w.r.t its aesthetics attributes in Section 3. Then we carry out the consistency analysis of the dataset in Section 4 to show the annotations are reliable that the raters have consistently labeled the images. Furthermore, in Section 6, we visually demonstrate the effectiveness of our model for aesthetics rating and analysis w.r.t the attributes. To address the effectiveness of content-aware model described in the paper, we analyze performance of different methods in utilizing this information in Section 5. Lastly, we attach the instruction used for teaching the raters to pass qualification test. The instruction and qualification test can filter out spammers to a large extent.