Giovani in un'ora - Ciclo di seminari - Terza parte
- Day - Time: 19 May 2022, h.11:00
Antonino Crivello - "Off-the-shelf Indoor localization Systems: is it feasible in the near future?"
Abstract: Localization systems are apparently close to hitting the market thanks to several companies offering off-the-shelf solutions to enable location-based services. Behind the curtains of marketing, several challenges must be addressed to reach the ambitious goal of wide-spreading indoor localization systems in heterogeneous environments. This talk will briefly introduce the big picture of the state of the art, including a discussion covering technological and methodological aspects. Then, the talk will cover the opportunities that FTM RTT may offer to indoor localization systems. Finally, the perspective will move from a single system to a system of systems to open to the need for interoperability mechanisms among different implementations of localization systems.
Nicola Messina - "Large-scale Image-Caption Matching using Cascaded Transformer-based Networks"
Abstract: With the increased accessibility of web and online encyclopedias, the amount of data to manage is constantly increasing. In Wikipedia, for example, there are millions of pages written in multiple languages. These pages contain images that often lack the textual context, remaining conceptually floating and therefore harder to find and manage.In this work, we present the system we designed for participating in the Wikipedia Image-Caption Matching challenge on Kaggle, whose objective is to use data associated with images (URLs and visual data) to find the correct caption among a large pool of available ones. A system able to perform this task would improve the accessibility and completeness of multimedia content on large online encyclopedias. Specifically, we propose a cascade of two models, both powered by the recent Transformer networks, able to efficiently and effectively infer a relevance score between the query image data and the captions. We verify through extensive experimentation that the proposed two-model approach is an effective way to handle a large pool of images and captions while maintaining bounded the overall computational complexity at inference time. We achieved the fifth position on the final private leaderboard, among more than 100 participating teams.
Giulio Ermanno Pibiri - "On Weighted K-Mer Dictionaries"
Abstract: We consider the problem of representing a set of k-mers (strings of length k over the DNA alphabet) and their abundance counts, or weights, in compressed space so that assessing membership and retrieving the weight of a k-mer is efficient. The representation is called a weighted dictionary of k-mers and finds application in numerous tasks in Bioinformatics that usually count k-mers as a pre-processing step. In fact, k-mer counting tools produce very large outputs that may result in a severe bottleneck for subsequent processing. In this short talk I will sketch the design of the SSHash index [Pibiri, ISMB (Bioinformatics journal) 2022] and illustrate its performance on large real-world datasets. Up to date, SSHash is the only k-mer dictionary that is exact, weighted, associative, fast, and small. The implementation is available at: https://github.com/jermp/sshash.