摘要
由一名新闻记者-机器人和机器学习的工作人员新闻编辑-每日新闻-关于机器学习的详细数据-在智能系统中已经呈现。根据NewsRx编辑在中国胶州的新闻报道，研究表明：“现有的基于Lea Rning的去雾算法由于缺乏配对的干净数据而难以处理真实世界的模糊图像。此外，大多数去雾方法需要大量的计算和内存。”本研究经费来源于国家自然科学基金(NSFC)。针对上述问题，本文提出了一种双教师知识边缘提取和无监督融合的单幅图像去雾方法，首先考虑到真实图像中复杂的退化因素，探索了两个合成到真实的去雾网络，采用非均质蒸馏策略生成两个初步的去雾结果。为了获得更合格的地面真值,提出了一种无监督对抗式融合网络k,对未配对干净图像的教师初始输出进行细化处理,特别是对未配对干净图像进行增强处理,并构造了一个中间图像来约束融合网络的输出。考虑到存储和存储开销,训练一个端到端的轻量级学生网络,学习从原始模糊图像到融合网络输出的映射.
Abstract
By a News Reporter-Staff News Editor at Robotics & Machine Learning Daily News Daily News-Data detailed on Machine Learning - In telligent Systems have been presented. According to news reporting out of Jiaozu o, People's Republic of China, by NewsRx editors, research stated, "Existing lea rning-based dehazing algorithms struggle to deal with real world hazy images for lack of paired clean data. Moreover, most dehazing methods require significant computation and memory." Financial support for this research came from National Natural Science Foundatio n of China (NSFC). Our news journalists obtained a quote from the research from HeNan Polytechnic U niversity, "To address the above problems, we propose a joint dual-teacher knowl edge distillation and unsupervised fusion framework for single image dehazing in this paper. First, considering the complex degradation factors in real-world ha zy images, two synthetic-to-real dehazing networks are explored to generate two preliminary dehazing results with the heterogeneous distillation strategy. Secon d, to get more qualified ground truth, an unsupervised adversarial fusion networ k is proposed to refine the preliminary outputs of teachers with unpaired clean images. In particular, the unpaired clean images are enhanced to deal with the d im artifacts. Furthermore, to alleviate the structure distortion in the unsuperv ised adversarial training, we constructed an intermediate image to constrain the output of the fusion network. Finally, considering the memory storage and compu tation overhead, an end-to-end lightweight student network is trained to learn t he mapping from the original hazy image to the output of the fusion network."
NETL
NSTL