Configuration
Color Scheme:
Output Layer Activation:
 
Layer Config:
 
Note: Adding many large layers will degrade performance.
 

View Mode: click an element above to see it enlarged.

*** To reset configuration and start again, please refresh the page.

SL Gfx Evolver

This is a simple application that uses the C64Style graphics library and very simple generated neural networks to produce images.

Quickstart

To begin using the default configuration, just click 'Start' above. A series of images will be generated using randomized networks. Clicking an image will produce new networks, similar to the one that generated the clicked image. With time and patience, you may get something interesting.

Notes

A population of networks is generated randomly to start, and one image is produced from each. When you click on one of the displayed images, the network that produced it is used to produce a new population. The new population of networks will be cloned from the selected one, and given random mutations to produce different looking images. The process then repeats when another image is selected. Over time, increasingly complex images can be "evolved".

Configuration

Use the form above to select the type of image output to generate, and the parameters for the networks. Click the '+' symbol to add additional 'hidden' layers to the networks. You can configure the number of nodes and the node activation type for each layer. When ready to generate the starting population, click 'Start'. Some configurations work better than others.

This is mostly just a toy, have fun! Enhancements may be coming in the future.

sl-gfx-evolver is (rarely) maintained by Shaun Lusk.
The source code for this project is freely available under the MIT license at Github.