Adversarial Appearance Learning in Augmented Cityscapes for Pedestrian Recognition in Autonomous Driving

2025-09-16

AI Summary

To address the domain shift between synthetic and real-world data that degrades pedestrian detection—particularly for vulnerable road users (VRUs)—in autonomous driving systems, this paper proposes a virtual pedestrian injection and augmentation framework tailored to the Cityscapes dataset. Methodologically, we design a novel adversarial generative network architecture that jointly models illumination, pose, and background interactions to synthesize photorealistic virtual pedestrians. We further integrate semantic and instance segmentation guidance with a virtual-real injection strategy to enhance geometric consistency and visual fidelity of the augmented scenes. Experimental results demonstrate that our approach significantly narrows the synthetic-to-real domain gap: it achieves up to a 4.2% mAP improvement over baseline methods on both pedestrian detection and segmentation tasks, thereby substantially improving model generalization to VRUs in real-world driving scenarios.

Technology Category

• Computer Vision: Computational Photography, Image & Video Synthesis
• Humans and AI: Game Design — Virtual Humans, NPCs and Autonomous Characters
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

In the autonomous driving area synthetic data is crucial for cover specific traffic scenarios which autonomous vehicle must handle. This data commonly introduces domain gap between synthetic and real domains. In this paper we deploy data augmentation to generate custom traffic scenarios with VRUs in order to improve pedestrian recognition. We provide a pipeline for augmentation of the Cityscapes dataset with virtual pedestrians. In order to improve augmentation realism of the pipeline we reveal a novel generative network architecture for adversarial learning of the data-set lighting conditions. We also evaluate our approach on the tasks of semantic and instance segmentation.

Problem

• Addressing domain gap between synthetic and real data
• Improving pedestrian recognition in autonomous driving
• Enhancing realism through adversarial lighting learning

Innovation

• Adversarial learning for lighting realism
• Augmenting Cityscapes with virtual pedestrians
• Pipeline for synthetic data generation