Current Deep Network (DN) visualization and interpretability methods rely heavily on data space visualizations such as scoring which dimensions of the data are responsible for their associated prediction or generating...
详细信息
ISBN:
(纸本)9798350301298
Current Deep Network (DN) visualization and interpretability methods rely heavily on data space visualizations such as scoring which dimensions of the data are responsible for their associated prediction or generating new data features or samples that best match a given DN unit or representation. In this paper, we go one step further by developing the first provably exact method for computing the geometry of a DN's mapping - including its decision boundary - over a specified region of the data space. By leveraging the theory of Continuous Piece-Wise Linear (CPWL) spline DNs, SplineCam exactly computes a DN's geometry without resorting to approximations such as sampling or architecture simplification. SplineCam applies to any DN architecture based on CPWL activation nonlinearities, including (leaky) ReLU, absolute value, maxout, and max-pooling and can also be applied to regression DNs such as implicit neural representations. Beyond decision boundary visualization and characterization, SplineCam enables one to compare architectures, measure generalizability, and sample from the decision boundary on or off the data manifold. Project website: ***/splinecam.
Due to the fact that a large number of honey bee hives have been lost in recent years, there have been some efforts to collect various data to assist with monitoring the hives and determine their health status. The Ap...
详细信息
ISBN:
(纸本)9781665476119
Due to the fact that a large number of honey bee hives have been lost in recent years, there have been some efforts to collect various data to assist with monitoring the hives and determine their health status. The Appalachian Multi-purpose Apiary Informatics Systems (AppMAIS) was funded by the North Carolina General Assembly in 2020 to establish a collaborative research project to monitor the health of honey bee hives. The AppMAIS Simple Data visualization App greatly increases our ability to analyze data, detect patterns, and train our ML swarm-like detection models. The app utilizes tooltips, hyperlinking, and highlight functionalities to significantly improve the efficiency of data analysis.
In this paper, we propose a large-capacity V2X communication data analysis and visualization tool for the development of high-safety/high-trust V2X communication systems, which are essential for Lv-4+ autonomous vehic...
详细信息
We present a feasibility-driven teleoperation framework designed to generate humanoid multi-contact maneuvers for use in unstructured environments. Our framework is designed for motions with arbitrary contact modes an...
详细信息
ISBN:
(纸本)9798350303278
We present a feasibility-driven teleoperation framework designed to generate humanoid multi-contact maneuvers for use in unstructured environments. Our framework is designed for motions with arbitrary contact modes and postures. The operator configures a pre-execution preview robot through contact points and kinematic tasks. A fast estimation of the preview robot's quasi-static feasibility is performed by checking contact stability and collisions along an interpolated trajectory. A visualization of Center of Mass (CoM) stability margin, based on friction and actuation constraints, is displayed and can be previewed if the operator chooses to add or remove contacts. Contact points can be placed anywhere on a mesh approximation of the robot surface, enabling motions with knee or forearm contacts. We demonstrate our approach in simulation and hardware on a NASA Valkyrie humanoid, focusing on multi-contact trajectories which are challenging to generate autonomously or through alternative teleoperation approaches.
Over the past thirty years, the rapid technological progress has sparked numerous breakthroughs in the area of the information and Communication Technologies. This has resulted in the creation of numerous engaging, in...
详细信息
Modern JavaScript development relies heavily on using Node Package Manager (NPM) modules. These modules are related by dependency relationships, possibly requiring dozens or hundreds of modules to build a complete Jav...
详细信息
ISBN:
(纸本)9798350308297
Modern JavaScript development relies heavily on using Node Package Manager (NPM) modules. These modules are related by dependency relationships, possibly requiring dozens or hundreds of modules to build a complete JavaScript web application. Studying dependencies, in terms of their sustainability, vulnerability, size, defects, etc., is fundamental for the deployment and maintenance of JavaScript web applications. We use a 3D metaphor based on presenting dependencies as an "elevated city", mapping both dependency relationships and characteristics of interest of each module. We developed a VR (virtual reality) scene representing the dependencies of several web applications using the elevated city metaphor, and exposed industrial experts to it to check its suitability. They explored a medium-sized project, with more than 200 dependencies, sharing their insights. The results highlight different aspects of our approach and how the combination of metrics helps experts to obtain insights from the ecosystem. The feedback shows the usefulness of the visualization to check and explore several aspects of the dependencies of an application, helping to identify problems related to maintainability, license usage, or vulnerabilities, and to design strategies to address them.
Maintaining large legacy systems often requires understanding their architecture. This is important since legacy system architecture decay over time and architecture violations may dramatically impact planned renovati...
详细信息
ISBN:
(纸本)9798350308297
Maintaining large legacy systems often requires understanding their architecture. This is important since legacy system architecture decay over time and architecture violations may dramatically impact planned renovation actions. Merely reading source files is time-consuming and often highly inefficient. visualizations have been proposed as a tool to support architecture understanding. Some software architecture visualizations decompose the software system architecture into layers, components, or slices from a structural viewpoint. Such visualizations, however, do not take into account the specificities of clientserver applications. They do not help maintainers identify and understand software architecture violations. In this paper, we propose CLISERVO, a new visualization to help software maintainers detect architectural violations in client-server systems. CLISERVO classifies client-server entities into different levels of dependencies, shared entities, or ambiguous entities (e.g., entities that belong abnormally to different layers). CLISERVO identifies and presents entities in their corresponding layers from two distinct viewpoints: global overview entities and violations, i.e ambiguous entities and illegal dependencies between layers. We validated our approach on three real-world industrial projects with access to their maintainers. We report the findings of 91 ambiguous entities, 29 purportedly shared and idle entities, 24 and 82 elements defined as shared but only used by the client or server, and 12 relations violating the layered architecture.
Depth perception is essential for our daily experiences, aiding in orientation and interaction with our surroundings. Virtual Reality allows us to decouple such depth cues mainly represented through binocular disparit...
详细信息
ISBN:
(纸本)9798350374490;9798350374506
Depth perception is essential for our daily experiences, aiding in orientation and interaction with our surroundings. Virtual Reality allows us to decouple such depth cues mainly represented through binocular disparity and motion parallax. Dealing with fully-mesh-based rendering methods these cues are not problematic as they originate from the object's underlying geometry. However, manipulating motion parallax, as seen in stereoscopic imposter-based rendering, raises multiple perceptual questions. Therefore, we conducted a user experiment to investigate how varying object sizes affect such visual errors and perceived 3-dimensionality, revealing an interestingly significant negative correlation and new assumptions about visual quality.
Automating apparel remanufacturing can improve textiles' ecological and societal impact, but integrating robotics into fashion is challenging due to fabric complexity. This demo presents a platform enabling non-ex...
详细信息
This paper serves as an introduction to the ECTS monitoring system and its potential applications in higher education. It also emphasizes the potential for ECTS monitoring to become a proactive system, supporting stud...
详细信息
ISBN:
(纸本)9798350394023;9798350394030
This paper serves as an introduction to the ECTS monitoring system and its potential applications in higher education. It also emphasizes the potential for ECTS monitoring to become a proactive system, supporting students by predicting academic success and identifying groups of potential dropouts for tailored support services. The use of the nearest neighbor analysis is suggested for improving data analysis and prediction accuracy.
暂无评论