EvoGAN: An evolutionary computation assisted GAN

The image synthesis technique is relatively well established which can generate facial images that are indistinguishable even by human beings. However, all of these approaches uses gradients to condition the output, resulting in outputting same with input. Also, they only basic expression or mimic an instead generating compound expression. In real life, however, expressions great diversity and complexity. this paper, we propose evolutionary algorithm (EA) assisted GAN, named EvoGAN, various any accurate target EvoGAN EA search results data distribution learned GAN. Specifically, use Facial Action Coding System (FACS) as encoding a pre-trained GAN images, then classifier recognize composition synthesized fitness function guide EA. Combined random searching algorithm, be easily sythesized. Quantitative Qualitative presented on several expressions, experimental demonstrate feasibility potential EvoGAN. source code available at https://github.com/ECNU-Cross-Innovation-Lab/EvoGAN.