Danya Yao

ASTAD: Asymmetric Style Transfer for Synthetic-to-Real Adaptation in Autonomous Driving

Dingyi Yao, Xinqi Zhang, Lihui Peng, Jianming Hu, Danya Yao, Yi Zhang

Synthetic data mitigates the data scarcity problem in autonomous driving perception. However, the synthetic-to-real gap leads to performance degradation, hindering real-world model generalization. Although current methods leverage diffusion models for photorealistic style transfe…Read more