š Hey Product Hunt! Weāve been working on this one for a while now and Iām really excited to share it with the community.
Over the last year weāve been experimenting with how machine learning can be incorporated into the artistic process. Weāre releasing the Face Maker AI as a free educational tool to demonstrate the potential of GANs.
Use the āPaletteā to pick a facial feature, like hair, skin, or glasses, draw on the sketch and watch AI imagine a face. If youāre on mobile press the āGenerate Faceā button. Supported keybindings: āBā - Brush, āIā - Eyedropper/Picker, ā [ ā - Decrease brush size, ā ] ā - Increase brush size, "Ctrl - Z" - Undo.
Please share your best and weirdest creations here!
This works by using a āsegmentation mapā. Deep Learning models are really good at performing per pixel labelling. Youāve probably seen the Remove Background tools that classify parts of the image either as background or foreground to make an alpha mask. But we donāt 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 āSegmentation 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 isnāt limited to just faces, you can do this with landscapes (like NVIDIAās GauGAN), buildings, cars, cats you name it!
Weāve added a free to use āsegment 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 š
This is for research or educational purposes only, but let me know if you have a commercial application and we can make it happen.
Massless Pen