I saw a repo related text2video and thought about customized porno generation. I made a plan including 3 stages:
1- make classifiers for ethnicity, age, actor attirbutes, position, and scenerio to label videos with timestamps.
The data creation method I used in stage one is download images from google,yandex, and bing and label them with search keyword. For example I searched for cowgirl porno and downloaded images and labeled respectively. (2000 images per class in total without augmentation)
I stuck at stage one because all of my models overfitting no matter what I tried. here is my email mertenes.yurtseven@gatech.edu for detailes.
I saw a repo related text2video and thought about customized porno generation. I made a plan including 3 stages: 1- make classifiers for ethnicity, age, actor attirbutes, position, and scenerio to label videos with timestamps. The data creation method I used in stage one is download images from google,yandex, and bing and label them with search keyword. For example I searched for cowgirl porno and downloaded images and labeled respectively. (2000 images per class in total without augmentation) I stuck at stage one because all of my models overfitting no matter what I tried. here is my email mertenes.yurtseven@gatech.edu for detailes.