Shoal

These images are generated by a neural network (Generative Adversarial Network - GAN) programmed in Python, with the Tensorflow library and deep learning-specific tools. The neural network is trained on hundreds of thousands of human-made images, employing techniques like transfer learning, style transfer, network distillation, conditional learning and distributed training. After being created, the images are passed through a second neural network (ESRGAN) that increases the resolution of the images.