DAMix: A Density-Aware Mixup Augmentation for Single Image Dehazing under Domain Shift
This addresses domain shift issues in image dehazing for computer vision applications, offering an incremental improvement in data efficiency and adaptivity.
The paper tackles performance degradation in single image dehazing under domain shift by proposing DAMix, a density-aware mixup augmentation that minimizes Wasserstein distance to target domain hazy images, achieving comprehensive domain adaptation improvements and enabling a network trained with half the source dataset using DAMix to outperform one trained with the full dataset without it.
Deep learning-based methods have achieved considerable success on single image dehazing in recent years. However, these methods are often subject to performance degradation when domain shifts are confronted. Specifically, haze density gaps exist among the existing datasets, often resulting in poor performance when these methods are tested across datasets. To address this issue, we propose a density-aware mixup augmentation (DAMix). DAMix generates samples in an attempt to minimize the Wasserstein distance with the hazy images in the target domain. These DAMix-ed samples not only mitigate domain gaps but are also proven to comply with the atmospheric scattering model. Thus, DAMix achieves comprehensive improvements on domain adaptation. Furthermore, we show that DAMix is helpful with respect to data efficiency. Specifically, a network trained with half of the source dataset using DAMix can achieve even better adaptivity than that trained with the whole source dataset but without DAMix.