CS236 GeoGAN: A Conditional GAN to Generate Maps from Satellite Images скачать в хорошем качестве

CS236 GeoGAN: A Conditional GAN to Generate Maps from Satellite Images

Automatically generating maps from satellite images is an important task. There is a body of literature which tries to address this challenge. We created a more expansive survey of the task by experimenting with different models and adding new losses to improve results. We created a database of pairs of satellite images and the corresponding map of the area. Then we tried to translate between the satellite images to the map images using three main model architectures: a conditional Generative Adversarial Network (GAN) which compresses the images down to a learned embedding, a generator which is trained as a normalizing flow model, and a conditional GAN which does not compress the original images. To improve the results we also added a reconstruction loss and style transfer loss in addition to the GAN losses. The third model architecture produced the best quality of sampled images. Since we have access to the real map for a given satellite image, we were able to assign a quantitative metric to the quality of the generated images in addition to inspecting them visually.