About

Leila De Floriani is a professor at the University of Maryland at College Park. She previously served as a professor at the University of Genova in Italy starting in 1990, where she developed the first undergraduate and graduate curricula in computer graphics in Italy and was the Director of the Ph.D. program in Computer Science for eight years. Throughout her career, she has held academic positions at the University of Nebraska, Rensselaer Polytechnic Institute, and the Italian National Research Council. De Floriani is recognized as a Fellow of IEEE, the International Association for Pattern Recognition (IAPR), and the Eurographics Association. She is also a Pioneer of the Solid Modeling Association and an inducted member of the IEEE Visualization Academy. She has received numerous honors including the IEEE Computer Society Golden Core award and membership in the IEEE Honor Society IEEE Eta Kappa Nu. She served as the 2020 President of the IEEE Computer Society and currently holds leadership roles within IEEE, including IEEE Division VIII Director for 2023-24 and membership on the IEEE Board of Directors and Assembly. Her research contributions span hierarchical models for graph representation and analysis, mesh-based multi-resolution modeling of surfaces and 3D shapes, level-of-detail modeling for analysis and visualization of 3D scalar fields, hierarchical terrain models, algorithms for visibility computation on triangulated terrain models, feature-based modeling for product design and manufacturing, data structures for meshes and simplicial complexes, and shape modeling and visualization based on combinatorial topology. Her current research focuses on spatial data analysis and visualization, spatial data structures, topological data analysis (TDA), and topology-based data visualization. De Floriani has authored over 300 peer-reviewed scientific publications in areas including data visualization, spatial data science, computer graphics, geometric modeling, and shape analysis, earning several best paper awards and keynote invitations. Her research has been funded by agencies such as the European Commission, the National Science Foundation, and NASA. De Floriani leads the UMD GeoVis group, which focuses on geospatial data representation and analysis. She has also been extensively involved in editorial and professional service roles, including editor-in-chief of the IEEE Transactions on Visualization and Computer Graphics from 2015 to 2018 and associate editor from 2004 to 2008. She currently serves as an editor for multiple journals including ACM Transactions on Spatial Algorithms and Systems, Computers & Graphics, Computer Science Review, GeoInformatica, Graphical Models, and the International Journal of Spatial Information Science. She has contributed to over 150 international conference program committees, often in leadership roles, and has been active in numerous IEEE committees and boards related to conferences, publications, and technical activities.

Research topics

• Computer Science
• Mathematics
• Artificial Intelligence
• Data Mining
• Geography
• Geometry
• Algorithm
• Theoretical computer science
• Optics
• Physics
• Remote sensing
• Combinatorics

Selected publications

• A Parallel Scale-Space Method for Critical Features Tracking on Triangulated Irregular Networks

  ACM Transactions on Spatial Algorithms and Systems · 2026-04-29

  articleSenior author

  The scale-space method is a well-established framework that constructs a hierarchical representation of an input signal and facilitates coarse-to-fine reasoning about its features. Considering the terrain elevation function as the input signal, the scale-space method can identify and track significant topographic features across different scales. The persistence of a feature, called its life span , indicates its importance and enables the automatic selection of critical features for applications such as cartography, nautical charting, and land-use planning. Traditional scale-space methods rely on gridded Digital Elevation Models (DEMs), which lack the flexibility to adapt to irregular input distributions and varied terrain complexity. In contrast, Triangulated Irregular Networks (TINs) can be directly generated from irregular point clouds and naturally preserve key features. In this work, we introduce a novel scale-space analysis pipeline for TINs, addressing the challenges in extending grid-based scale-space methods to irregular meshes. Building on our prior conference version, we further extend the pipeline in several directions: (i) a scale-aware, seam-free sampling strategy (S3) for high-quality TIN construction, (ii) an improved TIN smoothing method that combines virtual neighbors and angle re-weighting with a quantitative evaluation of geometric and gradient fidelity, and (iii) a fully parallel critical point tracking algorithm that eliminates global sorting and achieves substantial GPU speedups. Comprehensive experiments demonstrate that our TIN-based pipeline achieves superior geometric and topological fidelity, improved efficiency, and stronger resolution robustness than grid-based methods, making it a scalable and accurate framework for multi-scale terrain analysis.

  PublisherDOI

• ImplicitTerrainV2: Wavelet-Guided Spatially Adaptive Neural Terrain Representation

  arXiv (Cornell University) · 2026-05-21

  preprintOpen accessSenior author

  Digital elevation models (DEMs) underpin terrain analysis in Geographic Information Systems (GIS), but in their common raster form, they rely on interpolation for off-grid sampling and finite-difference operators for derivative-based analysis. Implicit neural representations (INRs) offer a continuous alternative, but prior terrain INRs lack explicit frequency control, neglect the gradient structure of terrain, and remain too large and costly to train for practical deployment. We present ImplicitTerrainV2, which advances terrain INRs toward a compact, efficient neural terrain data format by combining a spectral control mechanism with wavelet-guided spatial adaptivity, derivative-aware supervision, and post-training model compression. At its core, a wavelet complexity field (WCF) derives spatially-adaptive frequency masks from analytically computed wavelet coefficients, localizing high-frequency capacity to complex terrain regions. The same field guides complexity-aware adaptive sampling that concentrates training in high-complexity regions, while gradient matching applies extra supervision to enforce the smooth manifold structure of terrain DEMs for improved derivative fidelity. Post-training mixed-precision quantization and entropy coding reduce storage to 1.23 bpp with a 0.28 dB PSNR drop. On 50 Swiss terrain tiles, ImplicitTerrainV2 reaches 66.25 dB end-to-end PSNR, improving over the prior work by 5.70 dB while using 3.2x fewer parameters and training in 55 s per tile on a single GPU. Our compressed neural format is competitive with several established DEM codecs in rate-distortion performance, while additionally supporting off-grid point queries, closed-form derivative evaluation, and resolution-independent reconstruction, which may benefit many downstream GIS applications.

  PublisherDOI

• Extracting sea ice topographic features from triangulated surface models

  IEEE Transactions on Geoscience and Remote Sensing · 2026-01-01

  article

  Pressure ridges are the dominant topographic features of sea ice, playing a critical role in momentum transfer between the atmosphere and ocean. However, characterizing their three-dimensional morphology remains a challenge. Traditional methods, applied to linear profile data from laser altimeters, cannot extract two-dimensional features, specifically the horizontal orientation, extent, and connectivity of deformed ice. To address this gap, this paper introduces a novel strategy to extract pressure ridges directly from altimeter swath data by utilizing Triangulated Irregular Networks (TINs) and surface topology. Unlike raster grid-based models, TIN-based surface models do not require interpolation, thereby preserving the original point dataset and effectively retaining line features, which are essential for ridge extraction. Our algorithm integrates discrete Morse theory with surface roughness constraints to identify ridge structure lines from level sea ice. Additionally, we discuss how to address data dropout issues in the generation of TIN-based surface model, which prevents the correct extraction of ridge structures. The point cloud used in this study is high-resolution swath data from the NASA Operation IceBridge (OIB) Airborne Topographic Mapper (ATM). The validation against two independent datasets, the linear profile data collected by Ice, Cloud, and land Elevation Satellite (ICESat-2) and optical imagery captured during the OIB campaign, demonstrates that the proposed ridge extraction method can effectively detect ridge lines from TIN-based models. Such ridge features can provide more information of sea ice ridges, such as length, average height, and orientation, which are not possible when using linear profiles or gridded surface models.

  PublisherDOI

• ImplicitTerrainV2: Wavelet-Guided Spatially Adaptive Neural Terrain Representation

  ArXiv.org · 2026-05-21

  articleOpen accessSenior author

  Digital elevation models (DEMs) underpin terrain analysis in Geographic Information Systems (GIS), but in their common raster form, they rely on interpolation for off-grid sampling and finite-difference operators for derivative-based analysis. Implicit neural representations (INRs) offer a continuous alternative, but prior terrain INRs lack explicit frequency control, neglect the gradient structure of terrain, and remain too large and costly to train for practical deployment. We present ImplicitTerrainV2, which advances terrain INRs toward a compact, efficient neural terrain data format by combining a spectral control mechanism with wavelet-guided spatial adaptivity, derivative-aware supervision, and post-training model compression. At its core, a wavelet complexity field (WCF) derives spatially-adaptive frequency masks from analytically computed wavelet coefficients, localizing high-frequency capacity to complex terrain regions. The same field guides complexity-aware adaptive sampling that concentrates training in high-complexity regions, while gradient matching applies extra supervision to enforce the smooth manifold structure of terrain DEMs for improved derivative fidelity. Post-training mixed-precision quantization and entropy coding reduce storage to 1.23 bpp with a 0.28 dB PSNR drop. On 50 Swiss terrain tiles, ImplicitTerrainV2 reaches 66.25 dB end-to-end PSNR, improving over the prior work by 5.70 dB while using 3.2x fewer parameters and training in 55 s per tile on a single GPU. Our compressed neural format is competitive with several established DEM codecs in rate-distortion performance, while additionally supporting off-grid point queries, closed-form derivative evaluation, and resolution-independent reconstruction, which may benefit many downstream GIS applications.

  PublisherOA PDF

• Topology-based terrain segmentation using Apache Spark

  International Journal of Geographical Information Systems · 2025-10-30

  articleSenior author
  PublisherDOI

• SASNet: Spatially-Adaptive Sinusoidal Networks for INRs

  ArXiv.org · 2025-03-12

  preprintOpen accessSenior author

  Sinusoidal neural networks (SIRENs) are powerful implicit neural representations (INRs) for low-dimensional signals in vision and graphics. By encoding input coordinates with sinusoidal functions, they enable high-frequency image and surface reconstruction. However, training SIRENs is often unstable and highly sensitive to frequency initialization: small frequencies produce overly smooth reconstructions in detailed regions, whereas large ones introduce spurious high-frequency components that manifest as noise in smooth areas such as image backgrounds. To address these challenges, we propose SASNet, a Spatially-Adaptive Sinusoidal Network that couples a frozen frequency embedding layer, which explicitly fixes the network's frequency support, with jointly learned spatial masks that localize neuron influence across the domain. This pairing stabilizes optimization, sharpens edges, and suppresses noise in smooth areas. Experiments on 2D image and 3D volumetric data fitting as well as signed distance field (SDF) reconstruction benchmarks demonstrate that SASNet achieves faster convergence, superior reconstruction quality, and robust frequency localization -- assigning low- and high-frequency neurons to smooth and detailed regions respectively -- while maintaining parameter efficiency. Code available here: https://github.com/Fengyee/SASNet_inr.

  PublisherOA PDFDOI

• From Point Clouds to Forest Analytics: libTTS, a Practical Toolkit for the Geoindustry

  2025-11-03

  articleOpen accessSenior author

  Industries such as forestry, environmental consulting, and carbon credit markets increasingly benefit from laser scanning (LiDAR) to create a comprehensive forest inventory. These sectors require accurate, individual tree-level data for critical applications in ecological monitoring, carbon estimation, and resource management. However, a significant gap persists between academic research and practical, at-scale industrial deployment. Many current methods, including deep learning approaches, struggle with the reality of industrial settings, often requiring large, manually labeled datasets that are expensive to create and may not generalize well to new environments.

  PublisherOA PDFDOI

• Bathymetric mesh simplification for efficient two-dimensional barotropic tide modelling in New York Harbor

  Ocean Dynamics · 2025-11-05

  article
  PublisherDOI

• Sailing the Rough Seas of Automated Electronic Navigational Charts Compilation

  Advances in Cartography and GIScience of the ICA · 2025-10-20

  articleOpen access

  Abstract. Advancements in geospatial technology have benefited the hydrographic and maritime professions in many ways. Yet, compared to hydrographic data collection and processing, chart compilation workflows remain relatively slow, mainly due to limited human resources and the availability of automated algorithms that respect nautical charting constraints and Electronic Navigational Chart (ENC) database requirements. This work presents our research efforts to streamline the nautical chart compilation process through the introduction of automated processes and improving the efficiency and accuracy of existing. Among these processes are fundamental generalization tasks such as those for soundings, islands, and depth contours; ENC product specific requirements, such as those for reducing file size through the removal of collinear vertices forming polylines and polygons; and the updating of dependent features in the ENC database after generalization of one of their shared geometries.

  PublisherOA PDFDOI

• Critical Features Tracking on Triangulated Irregular Networks by a Scale-Space Method

  2024-10-29 · 5 citations

  preprintOpen accessSenior author

  The scale-space method is a well-established framework that constructs a hierarchical representation of an input signal and facilitates coarse-to-fine visual reasoning. Considering the terrain elevation function as the input signal, the scale-space method can identify and track significant topographic features across different scales. The number of scales a feature persists, called its life span, indicates the importance of that feature. In this way, important topographic features of a landscape can be selected, which are useful for many applications, including cartography, nautical charting, and land-use planning. The scale-space methods developed for terrain data use gridded Digital Elevation Models (DEMs) to represent the terrain. However, gridded DEMs lack the flexibility to adapt to the irregular distribution of input data and the varied topological complexity of different regions. Instead, Triangulated Irregular Networks (TINs) can be directly generated from irregularly distributed point clouds and accurately preserve important features. In this work, we introduce a novel scale-space analysis pipeline for TINs, addressing the multiple challenges in extending grid-based scale-space methods to TINs. Our pipeline can efficiently identify and track topologically important features on TINs. Moreover, it is capable of analyzing terrains with irregular boundaries, which poses challenges for grid-based methods. Comprehensive experiments show that, compared to grid-based methods, our TIN-based pipeline is more efficient, accurate, and has better resolution robustness.

  PublisherOA PDFDOI

Recent grants

• CGV: Small: Modeling and Analysis of Multidimensional Shapes

  NSF · $492k · 2011–2017

• III: Small: Geospatial Data Representation and Analysis through the Stellar Decomposition

  NSF · $500k · 2019–2024

• A Multi-resolution Approach to Modeling and Visualizing Multi-dimensional Scalar Fieldsu

  NSF · $300k · 2006–2011

Frequent coauthors

• Forrest Shull

  Fraunhofer USA Center Mid-Atlantic CMA

  391 shared

• Cecilia Metra

  Los Alamitos Medical Center

  329 shared

• David S. Ebert

  305 shared

• Riccardo Mariani

  301 shared

• Melissa Russell

  Los Alamitos Medical Center

  264 shared

• Anne Kelly

  255 shared

• Eric Berkowitz

  Los Alamitos Medical Center

  255 shared

• Saurabh Bagchi

  251 shared

Education

• Ph.D., Geography

  University of Maryland

  2000

• M.S., Geography

  University of Maryland

  1996

• B.A., Geography

  University of Rome 'La Sapienza'

  1993

Awards & honors

• Fellow of IEEE
• Fellow of the International Association for Pattern Recognit…
• Fellow of the Eurographics Association
• Pioneer of the Solid Modeling Association
• Inducted member of the IEEE Visualization Academy