arXiv subjects

Alexandros Haridis

Publications and source records attributed to Alexandros Haridis.

8 recordsLinked to original sources

arXiv2026

Rethinking Pretraining for Specialized Design Data: Evidence from the JONES-19 Cultural Design Dataset

Design and architectural archives encode expert human knowledge in graphical formats, providing a critical testbed for design-inspired Machine Learning (ML) challenges absent with typical computer vision benchmarks. Building on JONES-19, a small-size image dataset based on The Grammar of Ornament (London, 1857), we evaluate the discriminative performance of Convolutional Neural Networks (CNNs) in two model training strategies: (a) ImageNet pretraining for domain-general "visual common sense," and (b) learning from scratch on the design data in JONES-19. We find that while domain-general priors improve discriminative performance, learning from scratch augmented with repeated local sampling (multi-crop) effectively recovers these gains. For highly structured design data, local design-driven representations provide sufficient foundation for learning, challenging a reliance on massive general-purpose pretraining. These findings suggest that in specialized design domains, careful curation of smaller high-quality datasets that capture empirical and formal design principles may prove more effective and informative on the nature of a particular design domain than prioritizing large-scale data collection.

cs.LG↗

arXiv2025

Evaluation of Architectural Synthesis Using Generative AI

Recent advancements in multimodal Generative AI have the potential to democratize specialized architectural tasks, such as interpreting technical drawings and creating 3D CAD models, which traditionally require expert knowledge. This paper presents a comparative evaluation of two systems: GPT-4o and Claude 3.5, in the task of architectural 3D synthesis. We conduct a case study on two buildings from Palladio's Four Books of Architecture (1965): Villa Rotonda and Palazzo Porto. High-level architectural models and drawings of these buildings were prepared, inspired by Palladio's original texts and drawings. Through sequential text and image prompting, we assess the systems' abilities in (1) interpreting 2D and 3D representations of buildings from drawings, (2) encoding the buildings into a CAD software script, and (3) self-improving based on outputs. While both systems successfully generate individual parts, they struggle to accurately assemble these parts into the desired spatial relationships, with Claude 3.5 demonstrating better performance, particularly in self-correcting its output. This study contributes to ongoing research on benchmarking the strengths and weaknesses of off-the-shelf AI systems in performing intelligent human tasks that require discipline-specific knowledge. The findings highlight the potential of language-enabled AI systems to act as collaborative technical assistants in the architectural design process.

cs.AI↗

arXiv2024

Some Open Problems Regarding the Number of Lines and Slopes in Arrangements that Determine Shapes

A set $L$ of straight lines and a set $P$ of points in the Euclidean plane define an arrangement $\mathcal{A}$ = ($L$, $P$) of construction lines and registration marks, if and only if: (1) any point in $P$ is a point of intersection of at least two lines in $L$, and (2) any two nonparallel lines in $L$ have a unique point of intersection in $P$. This expository article discusses the following open problems regarding such point-line arrangements. Suppose $k \geq 0$ number of points are given in the plane. How many construction lines $k$ points must determine? How many distinct slopes, or directions, are defined by construction lines that $k$ points determine? How many distinct sets of construction lines partition the plane, such that the lines meet at exactly $k$ points? Empirical evidence is reported for small numbers of $k$, offering partial answers to the three problems. A conjecture is also stated for the first problem, on the number of construction lines, after examining a related problem about finite linear spaces from incidence geometry. This paper contributes to the body of work related to the mathematics of shapes in the area of shape grammar theory.

math.GM↗

arXiv2024

Geometry of Arrangements that Determine Shapes

Shape grammars compute over shapes which are defined in the universe $U^*$. Shapes in the universe $U^*$ are analogous to line drawings that can be physically realized in the plane. Any shape is embedded or contained in an arrangement of points and lines in the plane called, respectively, registration marks and construction lines, that satisfy special incidence laws. In this expository article, arrangements that contain shapes are studied as incidence structures and the finite geometries they give rise to are characterized. In particular, arrangements that contain shapes are distinguished into those that give rise to finite near-linear and linear spaces, and those that do not give rise to any proper form of geometry (in the strict mathematical sense). Arrangements that constitute finite geometries (near-linear and linear spaces) give an alternative characterization of determinate rules in shape grammars. This paper contributes to the body of work related to the mathematics of shapes in the area of shape grammar theory.

math.GM↗

arXiv2022

Analysis of shape grammars: continuity of rules

The rules in a shape grammar apply in terms of embedding to take advantage of the parts that emerge visually in the appearance of shapes. While the shapes are kept unanalyzed as a computation moves forward, part-structures for shapes can be defined retrospectively by analyzing how the rules were applied. An important outcome of this is that rule continuity is not builtin but it is "fabricated" retrospectively to analyze the computation as a continuous process. An aspect of continuity analysis that has not been addressed in the literature is how to decide which mapping forms to use to study the continuity of rule applications. This is addressed in this paper using recent results on shape topology and continuous mappings. A characterization is provided that distinguishes the suitable mapping forms from those that are inherently discontinuous or practically inconsequential for continuity analysis. It is also shown that certain intrinsic properties of shape topologies and continuous mappings provide an effective method of computing topologies algorithmically.

cs.FL↗

arXiv2022

Structure from Appearance: Topology with Shapes, without Points

A new methodological approach for the study of topology for shapes made of arrangements of lines, planes or solids is presented. Topologies for shapes are traditionally built on the classical theory of point-sets. In this paper, topologies are built with shapes, which are formalized without points, and with structures defined from their parts. An interpretative, aesthetic dimension is introduced according to which the topological structure of a shape is not inherited from an ambient space but is induced based on how its appearance is interpreted into sets of parts. The proposed approach provides a more natural, spatial framework for studies on the mathematical structure of design objects in art and design. More generally, it shows how mathematical constructs (here, topology) can be built directly in terms of objects of art and design, as opposed to the more common opposite approach where objects of art and design are subjugated to canonical mathematical constructs.

math.GN↗

arXiv2021

Natural Language and Spatial Rules

We develop a system that formally represents spatial semantics concepts within natural language descriptions of spatial arrangements. The system builds on a model of spatial semantics representation according to which words in a sentence are assigned spatial roles and the relations among these roles are represented with spatial relations. We combine our system with the shape grammar formalism that uses shape rules to generate languages (sets) of two-dimensional shapes. Our proposed system consists of pairs of shape rules and verbal rules where the verbal rules describe in English the action of the associated shape rule. We present various types of natural language descriptions of shapes that are successfully parsed by our system and we discuss open questions and challenges we see at the interface of language and perception.

cs.CL↗

arXiv2020

The Topology of Shapes Made with Points

In architecture, city planning, visual arts, and other design areas, shapes are often made with points, or with structural representations based on point-sets. Shapes made with points can be understood more generally as finite arrangements formed with elements (i.e. points) of the algebra of shapes $U_i$, for $i = 0$. This paper examines the kind of topology that is applicable to such shapes. From a mathematical standpoint, any "shape made with points" is equivalent to a finite space, so that topology on a shape made with points is no different than topology on a finite space: the study of topological structure naturally coincides with the study of preorder relations on the points of the shape. After establishing this fact, some connections between the topology of shapes made with points and the topology of "point-free" pictorial shapes (when $i > 0$) are discussed and the main differences between the two are summarized.

cs.GR↗