Mapping New Realities: Ground Truth Image Creation with Pix2Pix Image-to-Image Translation

TL;DR

This work tackles the scarcity of high-fidelity ground-truth imagery by applying Pix2Pix, a conditional GAN, to translate abstract maps into realistic aerial views. The authors design a Pix2Pix-based pipeline with a U‑Net generator and a PatchGAN discriminator, employing targeted data preprocessing and a robust training regimen to produce coherent ground-truth imagery. Key contributions include a paired map–aerial dataset, detailed architectural choices, and evidence of qualitative success in urban and rural contexts, suggesting wide applicability for urban planning and autonomous vehicle training. The approach promises a scalable method to generate realistic datasets, potentially accelerating geospatial analysis, simulation, and training pipelines where ground-truth data are scarce.

Abstract

Generative Adversarial Networks (GANs) have significantly advanced image processing, with Pix2Pix being a notable framework for image-to-image translation. This paper explores a novel application of Pix2Pix to transform abstract map images into realistic ground truth images, addressing the scarcity of such images crucial for domains like urban planning and autonomous vehicle training. We detail the Pix2Pix model's utilization for generating high-fidelity datasets, supported by a dataset of paired map and aerial images, and enhanced by a tailored training regimen. The results demonstrate the model's capability to accurately render complex urban features, establishing its efficacy and potential for broad real-world applications.

Figures (3)

• Figure 1: The Architecture of Generator. It is a U-Net-like architecture with an encoder-decoder structure.
• Figure 2: The Architecture of Discriminator
• Figure 3: Result Visualization