Abstract
Despite the remarkable achievement of recent underwater image restoration techniques, the lack of labeled data has become a major hurdle for further progress. In this work, we propose a mean-teacher based \textbf{Semi}-supervised \textbf{U}nderwater \textbf{I}mage \textbf{R}estoration (\textbf{Semi-UIR}) framework to incorporate the unlabeled data into network training. However, the naive mean-teacher method suffers from two main problems: (1) The consistency loss used in training might become ineffective when the teacher's prediction is wrong. (2) Using L1 distance may cause the network to overfit wrong labels, resulting in confirmation bias. To address the above problems, we first introduce a reliable bank to store the ``best-ever" outputs as pseudo ground truth. To assess the quality of outputs, we conduct an empirical analysis based on the monotonicity property to select the most trustworthy NR-IQA method. Besides, in view of the confirmation bias problem, we incorporate contrastive regularization to prevent the overfitting on wrong labels. Experimental results on both full-reference and non-reference underwater benchmarks demonstrate that our algorithm has obvious improvement over SOTA methods quantitatively and qualitatively. Code has been released at \href{this https URL}{this https URL}.
Abstract (translated)
尽管近年来水下图像恢复技术取得了显著成就,但缺乏标记数据已成为进一步进展的重大障碍。在本研究中,我们提出了一种基于平均教师(以下简称“教师”)的半监督水下图像修复(以下简称“半-UIR”)框架,将未标记数据纳入网络训练。然而,天真的平均教师方法面临两个主要问题:(1)在训练时,使用一致性损失可能会变得无效,如果教师的预测错误。(2)使用L1距离可能会导致网络过度适应错误的标签,导致确认偏差。为了解决上述问题,我们首先引入了一个可靠的存储库,以存储“最佳”输出作为伪 ground truth。为了评估输出质量,我们基于单调性特性进行了一项经验分析,以选择最可靠的 NR-IQA 方法。此外,考虑到确认偏差问题,我们引入了对比Regularization,以避免过度适应错误的标签。在全面参考和非参考水下基准测试中,实验结果证明,我们的算法在质量和数量上明显优于当前的最佳方法。代码已发布在 href{this https URL}{this https URL} 上。
URL
https://arxiv.org/abs/2303.09101
