Giovani in un'ora - Ciclo di seminari - Terza parte
- Day - Time: 30 November -0001, h.00:00
- Place: Area della Ricerca CNR di Pisa - Room: C-29
Alessio Ferrari - "Interview Review: Detecting Latent Ambiguities to Improve the Requirements Elicitation Process"
Abstract: Ambiguities identified during requirements elicitation interviews can be used by the requirements analyst as triggers for additional questions and, consequently, for disclosing further -- possibly tacit -- knowledge. Therefore, every unidentified ambiguity may be a missed opportunity to collect additional information. [Question/problem] Ambiguities are not always easy to recognize, especially during highly interactive activities such as requirements elicitation interviews. Moreover, since different persons can perceive ambiguous situations differently, the unique perspective of the interviewer might not be enough to identify all ambiguities. [Principal idea/results] To maximize the number of ambiguities recognized in interviews, this paper proposes a protocol to conduct reviews of requirements elicitation interviews. In the proposed protocol, the interviews are audio recorded and the recordings are inspected by both the interviewer and another analyst. The idea is to use the identified cases of ambiguity to create questions for the follow-up interviews. Our empirical evaluation of this protocol involves 42 students from Kennesaw State University and University of Technology Sydney. The study shows that, during the review, the interviewer and the other analyst identify 69\% of the total number of ambiguities discovered, while 31\% were identified during the interviews. Furthermore, the ambiguities identified by interviewers and other analysts during the review significantly differ from each other (a k-agreement of -0.46 is observed). [Contribution] Our results indicate that interview reviews allow the identification of a considerable number of undetected ambiguities, and can potentially be highly beneficial to discover unexpressed information in future interviews.
Manuele Sabbadin - "High Dynamic Range Point Clouds for Real-Time Relighting"
Abstract: Nowadays it is easy to acquire real environments using established 3D capture pipelines. The output of an acquisition comes in the form of a point cloud, which is a set of points representing the underlying surface. Each point sample often comes with additional attributes such as normal vector and color response. Although rendering and processing such data has been extensively studied, little attention has been devoted using the genuine light transport hidden in the recorded per-sample color response to relight virtual objects in visual effects (VFX) look-dev or augmented reality (AR) scenarios. In this seminar, I will illustrate a new framework which is able to relight a virtual object in real-time, exploiting the information recorded in an acquired point cloud. The framework is composed of two parts. First, since an acquired color point cloud typically comes in Low Dynamic Range (LDR) format, its range is expanded using a single HDR photo of the scene. Then, at rendering time, the resulting HDR point cloud is used to relight virtual objects, providing a diffuse model of the indirect illumination propagated by the environment. This is done with an ad-hoc version of the Point-Based Global Illumination algorithm which is able to relight the object in real-time. In the end, a couple of examples in an AR scenario will be provided.
Luca Pappalardo - "Towards a comprehensive data-driven evaluation of soccer players performance"
Abstract: The problem of evaluating the performance of soccer players is attracting the interest of many companies, websites, and the scientific community, thanks to the availability of massive data capturing many events generated during a game (e.g., tackles, passes, shots, etc.). Existing approaches are mainly mono-dimensional, in the sense that they propose a single metric capable of capturing just a single aspect of soccer performance. Recently, the PlayeRank algorithm has been proposed as a data-driven framework offering a principled multi-dimensional and role-aware evaluation of the performance of soccer players. We will show how to validate such a framework through an extensive experimental analysis advised by soccer experts, based on a massive dataset of millions of events pertaining five seasons of the prominent eleven soccer leagues in the world. Finally, we show how the PlayeRank framework can be used to characterize the typical performance of players and to predict their future performance on the basis of their performance history.