摘要:
1. Motivation and framework 当前基于CNN的高光谱图像修复取得了非常大的进展,但是仍然存在如下两个问题: the trained model is limited to the model-driven noise simulation process and may g 阅读全文
摘要:
代码地址:https://github.com/haoyuc/A2N 1、Motivation 注意力机制在超分辨率领域应用较多,作者提出两个问题: What feature will attention layer strengthen? Is attention always effective 阅读全文
摘要:
论文:Robust and generalizable visual representation learning via random convolutions, ICLR 2021 代码:https://github.com/tourdeml/robust-representation-ran 阅读全文
摘要:
论文:https://arxiv.org/pdf/2009.09238.pdf 代码:https://github.com/tsingqguo/efficientderain 对于图像去雨任务,作者指出当前方法的问题:1)the rain model assumptions adopted by m 阅读全文
摘要:
论文:https://arxiv.org/abs/2001.10291 代码:https://github.com/JimmyChame/SADNet 当前基于CNN的图像去噪方法取得很大进展,但是 Traditional CNN can use only the features in local 阅读全文