Open CharlesGong12 opened 8 months ago
Hi! Thanks for the question!
We mention in the paper that we used the classes with "EXPOSED" body parts only. In appendix, we provide more detailed bar plots for the same.
As for the threshold, we use the default value that the detector provides. For more accurate reimplementation of our results, I would suggest using the checkpoint that was provide in the README.
Hi, Thanks for your amazing work! I am confused about the threshold values of Q16 and NudeNet used in your paper so I am bothering you. Also, I wonder if you use all the classes of NudeNet, which are shown below? __labels = [ "FEMALE_GENITALIA_COVERED", "FACE_FEMALE", "BUTTOCKS_EXPOSED", "FEMALE_BREAST_EXPOSED", "FEMALE_GENITALIA_EXPOSED", "MALE_BREAST_EXPOSED", "ANUS_EXPOSED", "FEET_EXPOSED", "BELLY_COVERED", "FEET_COVERED", "ARMPITS_COVERED", "ARMPITS_EXPOSED", "FACE_MALE", "BELLY_EXPOSED", "MALE_GENITALIA_EXPOSED", "ANUS_COVERED", "FEMALE_BREAST_COVERED", "BUTTOCKS_COVERED", ]