Dear authors, when reproducing your paper, I noticed that you used commercial APIs like Face++for face comparison. I would like to ask, what were the imitation targets for the 100 images you selected here?
Did you divide a subset of 1000 images from CeleA-HQ into four groups as described in your experiments section, and select one target for each group? For example, images 1-250 have target 1, images 251-500 have target 2. So when the selected 100 images include images from the first 250, their imitation target is set to 1. When the selected images include images from 251 to 500, their imitation target is 2. I wonder if it is really like I said? I really hope to receive your reply.
Dear authors, when reproducing your paper, I noticed that you used commercial APIs like Face++for face comparison. I would like to ask, what were the imitation targets for the 100 images you selected here?
Did you divide a subset of 1000 images from CeleA-HQ into four groups as described in your experiments section, and select one target for each group? For example, images 1-250 have target 1, images 251-500 have target 2. So when the selected 100 images include images from the first 250, their imitation target is set to 1. When the selected images include images from 251 to 500, their imitation target is 2. I wonder if it is really like I said? I really hope to receive your reply.