An auto encoder network which gets generated circles as input and learns to decode new circles.
The project was created to get a better intuition on how auto encoders work. The gif file shows the decoded output after each epoch in the training process. So one can see how the network detects the relevant parameters in the bottleneck.