Multi-Agents GANs are a simple modification of the original architecture. Instead of one generator, there are multiple ones and they provide for a realistic variance in output. Each of the generators eventually collapse to different peaks of the distribution to be learned. Detailed information can be found in the report.
Project link: https://drive.google.com/file/d/1M72z3IaXSSF64P0Awr893FYFDfNYqUgm/view?usp=sharing