KLIEP-based Density Ratio Estimation for Semantically Consistent Synthetic to Real Images Adaptation in Urban Traffic Scenes

TL;DR

This paper introduces a KLIEP-based density ratio estimation method to improve semantic consistency in synthetic-to-real image translation for urban traffic scenes, enhancing the quality of images for autonomous driving tasks.

Contribution

The paper proposes a novel density prematching strategy using KLIEP to address semantic inconsistencies in adversarial image translation methods.

Findings

Improved semantic consistency in translated images.

Enhanced performance of semantic segmentation in autonomous driving.

Better image quality for downstream tasks.

Abstract

Synthetic data has been applied in many deep learning based computer vision tasks. Limited performance of algorithms trained solely on synthetic data has been approached with domain adaptation techniques such as the ones based on generative adversarial framework. We demonstrate how adversarial training alone can introduce semantic inconsistencies in translated images. To tackle this issue we propose density prematching strategy using KLIEP-based density ratio estimation procedure. Finally, we show that aforementioned strategy improves quality of translated images of underlying method and their usability for the semantic segmentation task in the context of autonomous driving.