Composition-Aware Scene Optimization for Product Images
This directory contains the supplemental material of submission "Composition-Aware Scene Optimization for Product Images".
This page consists of the comparison between the result with each term disabled and our result. The focus object(s) is specified in the parentheses.
This page consists of our optimization results after multiplying the weight of each energy term by 0.8 and 1.2, and the initial compositions of our optimization.
This page consists of the snapshots in Figure 2 in their original size.
This page consists of all 60 pair of detail images we generated from an overview by using our method and camera-only optimization. We generated detailed images for 20 randomly-picked objects in each of the three scenes: living room, study and kitchen. We then picked and presented 10 pairs to the participants in our user study, where the camera-optimization-only detail image and the full-optimization detail image differed the most. The 30 pairs included in the user study are listed first, and the remaining 30 examples are listed next. The rightmost column demonstrates the initial settings of our optimization and the camera-only optimization, which are results of a search along the canonical view direction.
This page consists of the retargeting results for different aspect ratios. Each cell consists of a comparison between our optimization (top) and the camera-only optimization (bottom). The results of retargeting from aspect ratio 'AxB' to aspect ratio 'CxD' is are in row 'AxB' and column 'CxD'. The input arrangement of each original aspect ratio (an optimal composition under the original aspect ratio) can be found in the cell on the diagonal.
This page consists of a sample webpage of our user study.
This page consists of tabulated study results and consistency analysis of non-experts session.