-
On the effectiveness of Rotation-Equivariance in U-Net: A Benchmark for Image Segmentation
Authors:
Robin Ghyselinck,
Valentin Delchevalerie,
Bruno Dumas,
Benoît Frénay
Abstract:
Numerous studies have recently focused on incorporating different variations of equivariance in Convolutional Neural Networks (CNNs). In particular, rotation-equivariance has gathered significant attention due to its relevance in many applications related to medical imaging, microscopic imaging, satellite imaging, industrial tasks, etc. While prior research has primarily focused on enhancing class…
▽ More
Numerous studies have recently focused on incorporating different variations of equivariance in Convolutional Neural Networks (CNNs). In particular, rotation-equivariance has gathered significant attention due to its relevance in many applications related to medical imaging, microscopic imaging, satellite imaging, industrial tasks, etc. While prior research has primarily focused on enhancing classification tasks with rotation equivariant CNNs, their impact on more complex architectures, such as U-Net for image segmentation, remains scarcely explored. Indeed, previous work interested in integrating rotation-equivariance into U-Net architecture have focused on solving specific applications with a limited scope. In contrast, this paper aims to provide a more exhaustive evaluation of rotation equivariant U-Net for image segmentation across a broader range of tasks. We benchmark their effectiveness against standard U-Net architectures, assessing improvements in terms of performance and sustainability (i.e., computational cost). Our evaluation focuses on datasets whose orientation of objects of interest is arbitrary in the image (e.g., Kvasir-SEG), but also on more standard segmentation datasets (such as COCO-Stuff) as to explore the wider applicability of rotation equivariance beyond tasks undoubtedly concerned by rotation equivariance. The main contribution of this work is to provide insights into the trade-offs and advantages of integrating rotation equivariance for segmentation tasks.
△ Less
Submitted 12 December, 2024;
originally announced December 2024.
-
Meaningful XAI Based on User-Centric Design Methodology
Authors:
Winston Maxwell,
Bruno Dumas
Abstract:
This report first takes stock of XAI-related requirements appearing in various EU directives, regulations, guidelines, and CJEU case law. This analysis of existing requirements will permit us to have a clearer vision of the purposes, the ``why'', of XAI, which we separate into five categories: contestability, empowerment/redressing information asymmetries, control over system performance, evaluati…
▽ More
This report first takes stock of XAI-related requirements appearing in various EU directives, regulations, guidelines, and CJEU case law. This analysis of existing requirements will permit us to have a clearer vision of the purposes, the ``why'', of XAI, which we separate into five categories: contestability, empowerment/redressing information asymmetries, control over system performance, evaluation of algorithmic decisions, and public administration transparency. The analysis of legal requirements also permits us to create four categories of recipients for explainability: data science teams; human operators of the system; persons affected by algorithmic decisions, and regulators/judges/auditors. Lastly, we identify four main operational contexts for explainability: XAI for the upstream design and testing phase; XAI for human-on-the-loop control; XAI for human-in-the-loop control; and XAI for ex-post challenges and investigations.Second, we will present user-centered design methodology, which takes the purposes, the recipients and the operational context into account in order to develop optimal XAI solutions.Third, we will suggest a methodology to permit suppliers and users of high-risk AI applications to propose local XAI solutions that are effective in the sense of being ``meaningful'', for example, useful in light of the operational, safety and fundamental rights contexts. The process used to develop these ``meaningful'' XAI solutions will be based on user-centric design principles examined in the second part.Fourth, we will suggest that the European Commission issue guidelines to provide a harmonised approach to defining ``meaningful'' explanations based on the purposes, audiences and operational contexts of AI systems. These guidelines would apply to the AI Act, but also to the other EU texts requiring explanations for algorithmic systems and results.
△ Less
Submitted 25 August, 2023;
originally announced August 2023.
-
A Metamodel and Prototype for Fluid Document Formats
Authors:
Ahmed A. O. Tayeh,
Bruno Dumas,
Beat Signer
Abstract:
With the transformation of computing from personal computers to the Internet, document formats have also seen some changes over the years. Future document formats are likely going to adapt to the emerging needs of ubiquitous computing, where information processing is embedded in everyday activities and objects. While most existing document formats have originally been a digital emulation of paper…
▽ More
With the transformation of computing from personal computers to the Internet, document formats have also seen some changes over the years. Future document formats are likely going to adapt to the emerging needs of ubiquitous computing, where information processing is embedded in everyday activities and objects. While most existing document formats have originally been a digital emulation of paper documents, over the years they have been enriched with additional digital features. These features were mainly incorporated to take advantage of the new functionality offered by the devices on which the documents are accessed. With the advent of ubiquitous computing, document formats seem to be facing the next evolutionary step. They will have to adapt to novel mobile devices, innovative interaction modalities, the distribution over multiple devices as well as heterogeneous input sources. This adaptation to the age of ubiquitous computing asks for several new document features. We outline a roadmap towards future fluid document representations for ubiquitous information environments. Based on the resource-selector-link (RSL) hypermedia metamodel - a general hypermedia metamodel supporting distribution, user rights management and content adaptation - we developed a metamodel for fluid document formats and the corresponding online text editor for fluid documents.
△ Less
Submitted 28 December, 2021;
originally announced December 2021.
-
Transitioning from Blackboard to Moodle amidst Natural Disaster: Faculty and Students Perceptions
Authors:
Ajayi Ekuase-Anwansedo,
Jose Noguera,
Brandon Dumas
Abstract:
Higher educational institutions continuously look for ways to improve the quality of their eLearning services and adapt learning solutions to suit the needs of the institution. During the 2016 Fall Semester, a university located in the Southern part of United States decided to transition from the Blackboard learning management system (LMS) to the Moodle learning management system. Typically such a…
▽ More
Higher educational institutions continuously look for ways to improve the quality of their eLearning services and adapt learning solutions to suit the needs of the institution. During the 2016 Fall Semester, a university located in the Southern part of United States decided to transition from the Blackboard learning management system (LMS) to the Moodle learning management system. Typically such a transition presents a huge challenge for the University staff, faculty, and students. Additionally, on August 2016, what CNN themedthe worst natural disaster, to strike the United States since Hurricane Sandy, occurred in Louisiana during the transition. This led to massive disruptions in activities throughout the state. This paper examines the perceptions of both faculty and student on the transition from one LMS to another and also what impact, if any, the natural disaster had on the process. Faculty and students were surveyed to gain understanding of how they perceived the transitioning process, their perception of both systems, their preferences, and why. Furthermore, we identified issues peculiar to transitioning during a natural disaster. The results of this study can be used to anticipate issues that may be associated with transitioning from one LMS to the other and issues peculiar to transitioning amidst a natural disaster. It can also be used to identify areas for improvement.
△ Less
Submitted 21 February, 2021;
originally announced February 2021.
