Growing Neural Cellular Automata Model

This is a model that learn to generate an image from one initial pixel. This model is based to the way that real multi-cellular organisms growth. Based to Growing Neural Cellular Automata paper.

Demo video

Requeriments

• anaconda

Setup

Step 1: Clone repository.

git clone https://github.com/adrianmarino/growing-neural-cellular-automata.git

Step 2: Create project environment.

conda env create --file environment.yml

See all ca-growth parameters

./ca-growth --help

optional arguments:
  -h, --help            show this help message and exit
  --config CONFIG       Configuration file name
  --action {train,test}
                        Specify train or test model
  --show-output         Show output evolution
  --hide-output         Hide output evolution
  --show-loss-graph     Show loss graph
  --hide-loss-graph     Hide loss graph

Test model

Step 1: Activate environment.

conda activate ca-growth

Step 1: Test an already trained config.

./ca-growth --action test --config lizard-16x16

Training a model

Step 1: Activate environment.

conda activate ca-growth

Step 2: Train model.

./ca-growth --action train --config lizard-16x16

Train in colab

Go to Growing_Neural_Cellular_Automata_Model notebook to train model under colab environment.