How deep learning extracts and learns leaf features for plant classification
How deep learning extracts and learns leaf features for plant classification
复制标题
DOI：
10.1016/j.patcog.2017.05.015
复制
发表时间：
2017-11-01
影响因子：
8
通讯作者：
Remagnino, Paolo
中科院分区：
文献类型：
--
作者：
Lee, Sue Han;Chan, Chee Seng;Remagnino, Paolo
对新兴网络（如软件定义网络（sdn））的攻击面进行系统的了解是必要的，并且可以说是使其更加安全的起点。以往的研究在很大程度上依赖于临时的经验方法，从不同的角度评估各种SDN元素的安全性。然而，他们还没有形成一种系统的方法，也没有开发出自动化的系统来严格测试sdn中的安全漏洞。因此，对新的SDN软件进行安全评估仍然是一个不可复制和不受管制的过程。本文对sdn中漏洞识别过程的自动化和标准化进行了研究。作为第一步，我们开发了一个安全评估框架DELTA，它可以在不同的测试环境中重新建立已发布的SDN攻击。接下来，我们使用协议感知模糊测试模块增强了我们的工具，以自动发现新的漏洞。在我们的评估中，DELTA成功地在不同的SDN控制器环境中重现了20种已知的攻击场景，并发现了7种新的SDN应用误导攻击。由计算机视觉研究人员开发的植物识别系统帮助植物学家更快地识别和鉴定未知的植物物种。迄今为止，许多研究都集中在最大化利用叶片数据库进行植物预测建模的程序或算法上，但这导致叶片特征容易随着不同的叶片数据和特征提取技术而变化。在本文中，我们使用卷积神经网络（CNN）直接从输入数据的原始表示中学习有用的叶子特征，并基于反卷积网络（DN）方法获得所选特征的直觉。我们报告了一些意想不到的结果：(1)与轮廓形状的特征相比，脉状的不同顺序是最具代表性的特征；(2)我们观察到叶片数据中的多层次表征，证明了特征从低级抽象到高级抽象的层次转换，对应于物种类别。我们发现这些发现符合叶性状的层次植物学定义。通过这些发现，我们获得了新的混合特征提取模型的设计见解，可以进一步提高植物分类系统的判别能力。源代码和模型可从https://github.comics-chan/Deep-Plant获得。(C) 2017 Elsevier Ltd.版权所有。
Developing a systematic understanding of the attack surface of emergent networks, such as software-defined networks (SDNs), is necessary and arguably the starting point toward making it more secure. Prior studies have largely relied on ad hoc empirical methods to evaluate the security of various SDN elements from different perspectives. However, they have stopped short of converging on a systematic methodology or developing automated systems to rigorously test for security flaws in SDNs. Thus, conducting security assessments of new SDN software remains a non-replicable and unregimented process. This paper makes the case for automating and standardizing the vulnerability identification process in SDNs. As a first step, we developed a security assessment framework, DELTA, that reinstantiates published SDN attacks in diverse test environments. Next, we enhanced our tool with a protocol-aware fuzzing module to automatically discover new vulnerabilities. In our evaluation, DELTA successfully reproduced 20 known attack scenarios across diverse SDN controller environments and discovered seven novel SDN application mislead attacks.Plant identification systems developed by computer vision researchers have helped botanists to recognize and identify unknown plant species more rapidly. Hitherto, numerous studies have focused on procedures or algorithms that maximize the use of leaf databases for plant predictive modeling, but this results in leaf features which are liable to change with different leaf data and feature extraction techniques. In this paper, we learn useful leaf features directly from the raw representations of input data using Convolutional Neural Networks (CNN), and gain intuition of the chosen features based on a Deconvolutional Network (DN) approach. We report somewhat unexpected results: (1) different orders of venation are the best representative features compared to those of outline shape, and (2) we observe multi-level representation in leaf data, demonstrating the hierarchical transformation of features from lower-level to higher-level abstraction, corresponding to species classes. We show that these findings fit with the hierarchical botanical definitions of leaf characters. Through these findings, we gained insights into the design of new hybrid feature extraction models which are able to further improve the discriminative power of plant classification systems. The source code and models are available at: https://github.comics-chan/Deep-Plant. (C) 2017 Elsevier Ltd. All rights reserved.