p/massless-pen
A pen for drawing in virtual reality
Jack Cohen

Face Maker AI ā€” Turn sketches into photorealistic faces

Featured
16
ā€¢
Face Maker AI takes a sketch and generates a photorealistic face
Replies
Best
Richard Casemore
I made a character for a trek story, with her mirror counter-part. Federation Terran
Dominic Hosler
@richard_casemore Cool! Looks a bit like commander T'Pol from Star Trek.
Jack Cohen
šŸ‘‹ Hey Product Hunt! Weā€™ve been working on this one for a while now and Iā€™m really excited to share it with the community. Over the last year weā€™ve been experimenting with how machine learning can be incorporated into the artistic process. Weā€™re releasing the Face Maker AI as a free educational tool to demonstrate the potential of GANs. Use the ā€œPaletteā€ to pick a facial feature, like hair, skin, or glasses, draw on the sketch and watch AI imagine a face. If youā€™re on mobile press the ā€œGenerate Faceā€ button. Supported keybindings: ā€œBā€ - Brush, ā€œIā€ - Eyedropper/Picker, ā€œ [ ā€œ - Decrease brush size, ā€œ ] ā€œ - Increase brush size, "Ctrl - Z" - Undo. Please share your best and weirdest creations here! This works by using a ā€œsegmentation mapā€. Deep Learning models are really good at performing per pixel labelling. Youā€™ve probably seen the Remove Background tools that classify parts of the image either as background or foreground to make an alpha mask. But we donā€™t have to stop there. We can classify an image with much more detail such as what parts contain an Eye, Ear, Nose, Mouthā€¦ etc. Passing an image of a face to a ML classification model can return a so-called ā€œSegmentation Mapā€, which assigns a unique color to each class of facial feature. Now this is where it gets interesting, we can play this in reverse, give the ML model a segmentation map, and use a GAN to generate the most convincing image of a face that it can (given the dataset it was trained on). The really cool thing is that this isnā€™t limited to just faces, you can do this with landscapes (like NVIDIAā€™s GauGAN), buildings, cars, cats you name it! Weā€™ve added a free to use ā€œsegment to faceā€ generator to our Media API and hooked it up to a basic paint interface in this tool. The model was trained using the Celeb HQ data set so you might see some hints of celebrities in the results šŸ˜‰ This is for research or educational purposes only, but let me know if you have a commercial application and we can make it happen.
Dominic Hosler
I'm not a fantastic artist, but I challenge any hairstylist to recreate this funky style! https://massless.media/FD87b8zKa...
Limited Timing
Love it! Great product! Thank you for creating such a cool product!
Dominic Hosler
@gcucci you're welcome! I'm glad you like it.
JoyQR
That's a fun product! Congrats on the launch
Jack Cohen
@joyqr Thank you! Glad we can bring a bit of joy šŸ˜Š
Abraham Yepremian
This is so cool!
Veronika Nesheva
Thanks so much @abeyep! Really glad you like it :)
Steven Nyland
This tool made my day. Splendid achievement!
Jack Cohen
@okavango Aww thanks so much, Steven šŸ™Œ