arXiv subjects

Tsung-Nan Lin

Publications and source records attributed to Tsung-Nan Lin.

4 recordsLinked to original sources

arXiv2022

DAMix: A Density-Aware Mixup Augmentation for Single Image Dehazing under Domain Shift

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.

cs.CV↗

arXiv2020

NTIRE 2020 Challenge on NonHomogeneous Dehazing

This paper reviews the NTIRE 2020 Challenge on NonHomogeneous Dehazing of images (restoration of rich details in hazy image). We focus on the proposed solutions and their results evaluated on NH-Haze, a novel dataset consisting of 55 pairs of real haze free and nonhomogeneous hazy images recorded outdoor. NH-Haze is the first realistic nonhomogeneous haze dataset that provides ground truth images. The nonhomogeneous haze has been produced using a professional haze generator that imitates the real conditions of haze scenes. 168 participants registered in the challenge and 27 teams competed in the final testing phase. The proposed solutions gauge the state-of-the-art in image dehazing.

cs.CV↗

arXiv2019

Guidelines for 5G End to End Architecture and Security Issues

Hackers target their attacks on the most vulnerable parts of a system. A system is therefore only as strong as its weakest part, similar to the Cannikin Law, which states that the capacity of a barrel of water depends on the height of the shortest rather than the longest piece of wood. To ensure the security of 5G networks, we first need to understand the overall 5G architecture and examine the potential threats instead of merely setting up a firewall. However, 5G networks will have tremendous coverage. The development of 5G techniques require extensive resources, leading to intense competition between countries attempting to develop 5G networks. Many outstanding papers discuss the techniques for developing specific aspects of the 5G architecture, but to the best of our knowledge, few provide an overview of a complete 5G network. This presents us with a difficult situation because we need to consider the overall architecture to ensure the security of 5G networks. To address that problem, in this paper we provide essential guidelines for understanding the architecture of 5G. We introduce 5G scenarios, outline the network architecture, and highlight the potential security issues for the various components of a 5G network. This paper is intended to facilitate a preliminary understanding of the 5G architecture and the possible security concerns of the various components. The end to end (E2E) 5G network architecture is composed of a next-generation radio access network (NG-RAN), multi-access edge computing (MEC), virtual evolved packet core (vEPC), a data network (DN) and a cloud service. Network slicing (NS), network function virtualization (NFV), NFV Management and Orchestration (MANO) and software-defined networking (SDN) are also important techniques for achieving 5G network architectures.

cs.NI↗

arXiv2019

Vanishing Nodes: Another Phenomenon That Makes Training Deep Neural Networks Difficult

It is well known that the problem of vanishing/exploding gradients is a challenge when training deep networks. In this paper, we describe another phenomenon, called vanishing nodes, that also increases the difficulty of training deep neural networks. As the depth of a neural network increases, the network's hidden nodes have more highly correlated behavior. This results in great similarities between these nodes. The redundancy of hidden nodes thus increases as the network becomes deeper. We call this problem vanishing nodes, and we propose the metric vanishing node indicator (VNI) for quantitatively measuring the degree of vanishing nodes. The VNI can be characterized by the network parameters, which is shown analytically to be proportional to the depth of the network and inversely proportional to the network width. The theoretical results show that the effective number of nodes vanishes to one when the VNI increases to one (its maximal value), and that vanishing/exploding gradients and vanishing nodes are two different challenges that increase the difficulty of training deep neural networks. The numerical results from the experiments suggest that the degree of vanishing nodes will become more evident during back-propagation training, and that when the VNI is equal to 1, the network cannot learn simple tasks (e.g. the XOR problem) even when the gradients are neither vanishing nor exploding. We refer to this kind of gradients as the walking dead gradients, which cannot help the network converge when having a relatively large enough scale. Finally, the experiments show that the likelihood of failed training increases as the depth of the network increases. The training will become much more difficult due to the lack of network representation capability.

cs.LG↗