Deep Face Drawing:

Made By:

Shu-Yu Chen  Wanchao Su  Lin Gao  Shihong Xia  Hongbo Fu

The DeepFaceDrawing project allows users with little to no training in drawing to produce high-quality face images from rough or even incomplete freehand sketches.

Fast generation of face images from freehand sketches is now possible using image-to-image deep translation techniques. There are several key issues solved by this project which were persistent in previous such deep face generating projects from sketches. One of the most prominent ones was the overfitting on sketches which often required professional sketches or even edge maps as input.

To solve this above mentioned issue and make the solution more viable and useful for people without any sketch drawing experience, the key approach used is to implicitly model the shape space of plausible face images and synthesize a realistic looking face image in this space to approximate an input sketch.

The project first aims to learn feature embeddings of key face components, and push corresponding parts of input sketches towards underlying component manifolds defined by the feature vectors of face component samples.

The reason for this project to work so well with even poor quality sketches is because it essentially uses input sketches as soft constraints and is thus able to produce high-quality face images even from rough and/or incomplete sketches.

Who can benefit from this project?

Artists and people without any training or experience in making sketches can benefit from this project for creating face images. The tool is easy to use even for non-artists, while still supporting fine-grained control of shape details.


Research information:

The Complete Research Paper is available here: Deep Face Drawing.

Train Your ML Models with Free GPU Computing Credits! 🙌

Try it yourself:

Play around and generate new face images with this Demo

Get your AI project featured

Fill out this Google Form and get a chance to be featured in this growing AI community

More Research Articles