Dataset Generation

Large Scale Dataset Generation for Face Recognition Applications.

Datasets

Download the files Texture Files, Segmentation Files, Light Files and put them in the root in their respective folders.

Models and Weights

Download the Models from the models folder and Weights from the weights folder and put them in the /models and /weights in root.

How to use

We can either generate with random attributes or manually generate an image.

Randomized Attributes

In generate.py in main generate() function on line 505, this will generate images from a combination of random attributes.

Controlled Attribute Generation

in generate.py in main, manual() function line 506 will lead to the manual function in line 462. This is an example code of how to generate an image given constraints; this can be modifiable to be able to read files for generation.

Process Visualization

Texture Infusion -> Inversion -> Relighting -> Pose Transfer -> Image Enhancement -> Age Manipulation

Other Examples

Custom Attribute Manipulation

Random Attribute Generation

Variety of Different Identities

Variety of Same Identities

Repositories Used

SPADE, Inversion, Relighting, Pose Transfer, Image Enhancement, Age Manipulation