当前位置:
X-MOL 学术
›
arXiv.cs.CV
›
论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Streaming Networks: Increase Noise Robustness and Filter Diversity via Hard-wired and Input-induced Sparsity
arXiv - CS - Computer Vision and Pattern Recognition Pub Date : 2020-03-30 , DOI: arxiv-2004.03334 Sergey Tarasenko and Fumihiko Takahashi
arXiv - CS - Computer Vision and Pattern Recognition Pub Date : 2020-03-30 , DOI: arxiv-2004.03334 Sergey Tarasenko and Fumihiko Takahashi
The CNNs have achieved a state-of-the-art performance in many applications.
Recent studies illustrate that CNN's recognition accuracy drops drastically if
images are noise corrupted. We focus on the problem of robust recognition
accuracy of noise-corrupted images. We introduce a novel network architecture
called Streaming Networks. Each stream is taking a certain intensity slice of
the original image as an input, and stream parameters are trained
independently. We use network capacity, hard-wired and input-induced sparsity
as the dimensions for experiments. The results indicate that only the presence
of both hard-wired and input-induces sparsity enables robust noisy image
recognition. Streaming Nets is the only architecture which has both types of
sparsity and exhibits higher robustness to noise. Finally, to illustrate
increase in filter diversity we illustrate that a distribution of filter
weights of the first conv layer gradually approaches uniform distribution as
the degree of hard-wired and domain-induced sparsity and capacities increases.
中文翻译:
流媒体网络:通过硬连线和输入引起的稀疏性提高噪声鲁棒性和滤波器多样性
CNN 在许多应用中都取得了最先进的性能。最近的研究表明,如果图像被噪声破坏,CNN 的识别精度会急剧下降。我们专注于噪声损坏图像的鲁棒识别精度问题。我们介绍了一种称为流媒体网络的新型网络架构。每个流都以原始图像的某个强度切片作为输入,并且独立训练流参数。我们使用网络容量、硬连线和输入引起的稀疏性作为实验的维度。结果表明,只有同时存在硬连线和输入诱导稀疏性才能实现稳健的噪声图像识别。Streaming Nets 是唯一具有两种类型的稀疏性并且对噪声具有更高鲁棒性的架构。最后,
更新日期:2020-04-10
中文翻译:
流媒体网络:通过硬连线和输入引起的稀疏性提高噪声鲁棒性和滤波器多样性
CNN 在许多应用中都取得了最先进的性能。最近的研究表明,如果图像被噪声破坏,CNN 的识别精度会急剧下降。我们专注于噪声损坏图像的鲁棒识别精度问题。我们介绍了一种称为流媒体网络的新型网络架构。每个流都以原始图像的某个强度切片作为输入,并且独立训练流参数。我们使用网络容量、硬连线和输入引起的稀疏性作为实验的维度。结果表明,只有同时存在硬连线和输入诱导稀疏性才能实现稳健的噪声图像识别。Streaming Nets 是唯一具有两种类型的稀疏性并且对噪声具有更高鲁棒性的架构。最后,