Web template system | EPFL Graph Search

A web template system in web publishing allows web designers and developers work with web templates to automatically generate custom web pages, such as the results from a search. This reuses static web page elements while defining dynamic elements based on web request parameters. Web templates support static content, providing basic structure and appearance. Developers can implement templates from content management systems, web application frameworks, and HTML editors. A web template system is composed of the following: A template engine: the primary processing element of the system; Content resource: any of various kinds of input data streams, such as from a relational database, XML files, LDAP directory, and other kinds of local or networked data; Template resource: web templates specified according to a template language; The template and content resources are processed and combined by the template engine to mass-produce web documents. For purposes of this article, web documents include any of various output formats for transmission over the web via HTTP, HTTPS, or another Internet protocol. With the model typically held in a relational database, the remaining components of the MVC architecture are the control and view. In the simplest of systems these two are not separated. However, adapting the separation of concerns principle one can completely decouple the relationships. For example, the view template may look like this: Sites Then, the control template loads the view, and then uses XPath addressing to insert components from a database, for instance: preserveWhiteSpace = false;

doc->Load('view.html');

titlenode =

doc->createTextNode("Like this");

xpath = new DOMXPath(

doc);

xpath->registerNamespace("h", "http://www.w3.

Understanding Deep Neural Networks using Adversarial Attacks

Krishna Kanth Nakka

Deep Neural Networks (DNNs) have achieved great success in a wide range of applications, such as image recognition, object detection, and semantic segmentation. Even thoughthe discriminative power of DNNs is nowadays unquestionable, serious concerns have arised ever since DNNs have shown to be vulnerable to adversarial examples craftedby adding imperceptible perturbations to clean images. The implications of these malicious attacks are even more significant for DNNs deployed in real-world systems, e.g.,autonomous driving and biometric authentication. Consequently, an intriguing question that we aim to understand is the underlying behavior of DNNs to adversarial attacks.This thesis contributes to a better understanding of the mechanism of adversarial attacks on DNNs. Our main contributions are broadly in two directions: (1) we proposeinterpretable architectures first to understand the reasons for the success of adversarial attacks and then to improve the robustness of DNNs; (2) we design intuitive adversarialattacks to both mislead and use as a tool to expand our present understanding of DNNs' internal workings and their limitations. In the first direction, we introduce deep architectures that allow humans to interpret the reasoning process of DNNs prediction. Specifically, we incorporate Bag-of-visual-wordsrepresentations from the pre-deep learning era into DNNs using an attention scheme. We find key reasons for adversarial attack success and use these insights to propose anadversarial defense by maximally separating the latent features of discriminative regions while minimizing the contribution of non-discriminative regions in the final prediction.The second direction deals with the design of adversarial attacks to understand DNNs' limitations in a real-world environment. To begin with, we show that existing state-of-the-art semantic segmentation networks that achieve superior performance by exploiting the context are highly susceptible to indirect local attacks. Furthermore, we demonstrate the existence of universal directional perturbations that are quasi-independent of the input template but still successfully fool unknown siamese-based visual object trackers. We then identify that the mid-level filter banks across different backbones bear strong similarities and thus can be potential common ground for attack. We, therefore, learn a generator that disrupts mid-level features with high transferability across different target architectures, datasets, and tasks. In short, our attacks highlight critical vulnerabilities of DNNs, which make their deployment challenging in the real-world environment, even in the extreme case when the attacker is unaware of the target architecture or the targetdata used to train it.Furthermore, we go beyond fooling networks and demonstrate the usefulness of adversarial attacks for studying the internal disentangled representations in self-supervised 3D pose estimation networks. We observe that adversarial manipulation of appearance information in the input image alters the pose output, indicating that the pose code contains appearance information and disentanglement is far from complete. Besides the above contributions, an underlying theme that arises multiple times in this thesis is counteracting the adversarial attacks by detecting them.

EPFL2022

Template-assisted electrodeposition of nickel and nickel copper 3D microcomponents

Patrik Schürch

New types of metamaterials and architectured material require metallic materials with precise structural design in the microscale. However, additivemanufacturing of metallic structures in themicroscale has proven difficult, as the demand for the material quality and accuracy of the 3D structure is high. Of themultitude of techniques available, template-assisted electrodeposition (TAE), also known by its German abbreviation, LIGA, has shown excellent results for the production of 2D components. With new lithography techniques, especially two-photon lithography (TPL), TAE shows promise of microscale additive manufacturing for 3D metal structures. This thesis aims to investigate the electrodeposition into a 3D template and discuss the influence of the template on the local current density and the resulting microstructure of the deposit. 3D finite element simulations were used to understand the process of template filling. The electrodeposited structures were investigated towards their shape accuracy as well as their microstructure. Various deposited structures were used for micromechanical investigation with the explicit regard to inspect their properties and performance for future use in MEMS-devices. Simple pillars were investigated at first to inspect themechanical properties and arrangement of the templates in arrays. In this work, subsequently, 3D micro springs were produced and their properties compared with theoretical mechanical simulations. These springs exhibited significant strength, unprecedented in microsprings of this size. Following simple templates, more complex templates with multiple ongoing growth fronts were investigated. Micromechanical shear test specimen were produced and the template design was investigated. These templates were also used to investigate the effectiveness of pulse electrodeposition was investigated to fill corners. Furthermore, the microstructure showed elongated grains along the growth direction allowing insight into the deposits growth. Where to growth fronts meet an accumulation of grain boundaries was visible in the microstructure, but showed no porosity. Building on these results, microlattices with a body center cubic structure were designed and tested. The microlattices showed remarkably high strength as well as excellent promise for energy absorption. Building on the success of the technique and the previous structures, the electrodeposition of Copper Nickel alloys was investigated. A 3D time-dependent pulse electrodeposition simulation was created and the deposition of large 2D components was simulated and compared to experimental values. Overall, trough these specific case studies, it was shown that TAE combined with TPL produces microcomponents with excellent accuracy, homogenous microstructures with no significant defects and excellent mechanical properties. The method has shown reproducibility. The insight gained by the electrodeposition simulation can be used in the future for others to emulate the process or create their electrodeposition simulation.