Mutual learning for domain adaptation: Self-distillation image dehazing network with sample-cycle

Deep learning-based methods have made significant achievements for image dehazing. However, most of existing dehazing networks focus on training models using synthetic hazy images, resulting in generalization performance degradation when applying on real-world hazy images because of the domain shift problem. In this paper, we propose a mutual learning dehazing framework for domain adaption. Specifically, we first devise two siamese networks: a teacher network in the synthetic domain and a student network in the real domain, and then optimize them in a mutual learning manner by leveraging the exponential moving average (EMA) and joint loss. Moreover, we design a sample-cycle strategy based on haze density augmentation (HDA) module to introduce pseudo real-world image pairs provided by the student network into training for further improving the generalization performance. Extensive experiments demonstrate that the proposed framework outperforms state-of-the-art dehazing techniques in terms of subjective and objective evaluation.