A Shared Representation for Photorealistic Driving Simulators

TL;DR

This paper introduces a semantically-aware discriminator architecture for cGANs that enhances photorealistic image synthesis in driving simulators, improving image quality across various datasets and applications.

Contribution

The authors propose a novel discriminator design that learns a shared latent space for semantic segmentation and content reconstruction, improving image realism in conditional image synthesis.

Findings

Enhanced image quality in driving simulation datasets

Applicable to scene, building, and human synthesis tasks

Improved guidance for generator in cGANs

Abstract

A powerful simulator highly decreases the need for real-world tests when training and evaluating autonomous vehicles. Data-driven simulators flourished with the recent advancement of conditional Generative Adversarial Networks (cGANs), providing high-fidelity images. The main challenge is synthesizing photorealistic images while following given constraints. In this work, we propose to improve the quality of generated images by rethinking the discriminator architecture. The focus is on the class of problems where images are generated given semantic inputs, such as scene segmentation maps or human body poses. We build on successful cGAN models to propose a new semantically-aware discriminator that better guides the generator. We aim to learn a shared latent representation that encodes enough information to jointly do semantic segmentation, content reconstruction, along with a coarse-to-fine grained adversarial reasoning. The achieved improvements are generic and simple enough to be applied to any architecture of conditional image synthesis. We demonstrate the strength of our method on the scene, building, and human synthesis tasks across three different datasets. The code is available at https://github.com/vita-epfl/SemDisc.