Ioannis Dravilas

Mr. Ioannis Dravilas

Informatics and Telecommunications, National and Kapodistrian University of Athens, Athens, Attikí, Greece
orcid

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
Verified
Conference Paper
Open access
PDF file
Ioannis DravilasIoannis Dravilas
Ioannis Dravilas
Stavros DafisStavros Dafis
Stavros Dafis
Giorgos KyrosGiorgos Kyros
Giorgos Kyros
KLKonstantinos Lagouvardos
Konstantinos Lagouvardos
MKManolis Koubarakis
Manolis Koubarakis
Remote sensing of snowfall has been proved to be a great challenge since the start of the satellite era. Several techniques have been applied to satellite data to estimate the fraction of frozen precipitation that reaches the surface. This study aims at investigating the efficacy of machine learning (ML), and especially deep learning (DL), in estimating the precipitation phase of the Integrated Multi-satellitE Retrievals for the Global Precipitation Measurement (GPM-IMERG). To achieve this, a training phase with hourly high-resolution numerical model outputs and in situ data was chosen for the period of late-2020 and 2021. Preliminary results show that ML models can estimate the precipitation phase with relatively high accuracy based on several case studies. The findings suggest that ML models offer a promising approach for advancing the nowcasting of snowfall and building a long-term archive dataset of IMERG-based snowfall using conventional real-time data.


13 February 2026 in arXiv
Verified
Preprint
Open access
PDF file
Ioannis DravilasIoannis Dravilas
Ioannis Dravilas
IKIoannis Kapetangeorgis
Ioannis Kapetangeorgis
ALAnastasios Latsoudis
Anastasios Latsoudis
CMConor McCarthy
Conor McCarthy
GMGonçalo Marcelino
Gonçalo Marcelino
MWMarcel Worring
Marcel Worring
Composed Image Retrieval (CIR) allows users to search for images by combining a reference image with a text prompt that describes desired modifications. While vision-language models like CLIP have popularized this task by embedding multiple modalities into a joint space, developers still lack tools that reveal how these multimodal prompts interact with embedding spaces and why small wording changes can dramatically alter the results. We present InfoCIR, a visual analytics system that closes this gap by coupling retrieval, explainability, and prompt engineering in a single, interactive dashboard. InfoCIR integrates a state-of-the-art CIR back-end (SEARLE arXiv:2303.15247) with a six-panel interface that (i) lets users compose image + text queries, (ii) projects the top-k results into a low-dimensional space using Uniform Manifold Approximation and Projection (UMAP) for spatial reasoning, (iii) overlays similarity-based saliency maps and gradient-derived token-attribution bars for local explanation, and (iv) employs an LLM-powered prompt enhancer that generates counterfactual variants and visualizes how these changes affect the ranking of user-selected target images. A modular architecture built on Plotly-Dash allows new models, datasets, and attribution methods to be plugged in with minimal effort. We argue that InfoCIR helps diagnose retrieval failures, guides prompt enhancement, and accelerates insight generation during model development. All source code allowing for a reproducible demo is available at https://github.com/giannhskp/InfoCIR.

20 January 2025
Verified
Conference Paper
Open access
SKStavros Keppas
Stavros Keppas
HBHaris Balis
Haris Balis
Ioannis DravilasIoannis Dravilas
Ioannis Dravilas
JPJohn Pagonis
John Pagonis
Designing a weather station network is a demanding, multi-objective optimisation problem and usually constrained to local geographies. In this study, the authors deviate from typical approaches that focus the design of weather station networks on a small or country-wide area and present a method that is applicable on a global scale.Prior art suggests that weather networks should exhibit high density, often at 1-3km or finer resolution, especially when deployed over complex topographies and urban landscapes. High station density is usually required to support research on urban micrometeorology, agricultural applications and to capture intricate meteorological mesoscale phenomena such as convective precipitation and sea breeze. High density is also required due to the persistence of temperature inversions at near-surface layers is significantly influenced by topography, leading to prolonged periods of temperature inversion.In this novel approach, the authors suggest the design of a global weather network distributed over millions of hexagons covering the entire world. The number of weather stations per hexagon is determined by the topology (e.g. maximum elevation difference, aspects, water formations, etc.) and the land use (urban coverage, green areas, etc.) of the covered area.The method is materialized via an open-source software tool (available on GitHub) which utilizes freely available elevation data (Copernicus DEM) and land use data (OpenStreetMap) and is capable of preparing the global weather station network in reasonable computation time (~24 hours on a 16-core CPU).Finally, the authors present their findings, discuss the effect of various hexagon sizes and suggest that the design of a global weather station network is viable and computationally feasible.