Fig. 6From: Image embedding and user multi-preference modeling for data collection samplingColumns 1–5 represent, respectively, the user-preferred category in the coco dataset, the t-SNE visualization for all images embedded in the feature space, sampled results for sample size, M=500, sampled results for sample size, M=1000 and sampled results for sample size, M=1500 for the proposed method in row 1, 3 and 5 and for the random sampling method in rows 2, 4 and 5Back to article page