Dok11 / nn-gan-panoramo

Implementation of Generative Adversarial Network for equirectangular panorama generation

nn-gan-panoramo

Implementation of Generative Adversarial Network for equirectangular panorama generation

Project structure

/data/train*.npz - File with Numpy array of images with 128x64x1 dimension
/images - Presentation of images with training result at several epochs
/train-images - Dataset with panorama images (jpg) with 640x320x3
/gan.py - GAN based on dense layers for generation panorama images 128x64x1
/gan_src.py - GAN based on dense layers for MNIST digits generation 28x28x1. Copy from https://raw.githubusercontent.com/eriklindernoren/Keras-GAN/master/gan/gan.py
/prepare-data.py - Script for creation npz file with images from /images