Since GAN's are totally supervised learning, It is totally on the author or programmer of the GAN to pick the CNN for it. The CNN could be a birds or a morph between time in a local area, or sequence through time. When the generator NN is generating something between a chicken and a turkey the CNN fire 0.5 for chicken and 0.5 for turkey. A 1.0 means it is a chicken on the chicken output one the the other side of the CNN. Or 1.0 for turkey for the dedicated output for turkey.
The generator NN has a value you input for chicken, let say the value five. And turkey has a value of eight. You put a the value of five into the input of the generator network and image of a chicken pop out on the other side and the CNN. That activates it dedicated chicken output when it see the chicken appear on the output of the generator NN. A six inputted into the generator NN will generate some thing in between the two birds. The CNN will activate a 0.5 on the chicken line and 0.5 on the turkey line. You need a register for each morph, hair, color, and etc..... Which are inputted into the front end of the Generator NN. The training algorithm is needed for the generator NN. Announcement: Face Editor Download: https://www.youtube.com/watch?v=p7k86wwUoJg ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Td13181dbab529095-M736cd50607290d593080b176 Delivery options: https://agi.topicbox.com/groups/agi/subscription
