Muhammad Tayyab Khan, Zane Yong, Lequn Chen
et al.
Engineering drawings are fundamental to manufacturing communication, serving as the primary medium for conveying design intent, tolerances, and production details. However, interpreting complex multi-view drawings with dense annotations remains challenging using manual methods, generic optical character recognition (OCR) systems, or traditional deep learning approaches, due to varied layouts, orientations, and mixed symbolic-textual content. To address these challenges, this paper proposes a three-stage hybrid framework for the automated interpretation of 2D multi-view engineering drawings using modern detection and vision language models (VLMs). In the first stage, YOLOv11-det performs layout segmentation to localize key regions such as views, title blocks, and notes. The second stage uses YOLOv11-obb for orientation-aware, fine-grained detection of annotations, including measures, GD&T symbols, and surface roughness indicators. The third stage employs two Donut-based, OCR-free VLMs for semantic content parsing: the Alphabetical VLM extracts textual and categorical information from title blocks and notes, while the Numerical VLM interprets quantitative data such as measures, GD&T frames, and surface roughness. Two specialized datasets were developed to ensure robustness and generalization: 1,000 drawings for layout detection and 1,406 for annotation-level training. The Alphabetical VLM achieved an overall F1 score of 0.672, while the Numerical VLM reached 0.963, demonstrating strong performance in textual and quantitative interpretation, respectively. The unified JSON output enables seamless integration with CAD and manufacturing databases, providing a scalable solution for intelligent engineering drawing analysis.
The purpose of the research is to comprehensively analyze the use of botanical illustration in graphic design and to identify innovative methods for its creation. Research methods. To identify and systematize current trends in the use of botanical illustration within graphic design, general scientific methods such as analysis, synthesis, deduction, comparison, generalization, observation, and interdisciplinary synthesis were applied. Information from the study of design projects was organized using the grouping method. Scientific novelty. As a result of examining various approaches to the creation and practical application of botanical illustration, it is proposed for the first time to distinguish between several types of graphic reproduction: botanical illustration, botanical motifs, and digital botanical illustration. Conclusions. The results of the study show that botanical illustration requires deeper analysis within the context of graphic design and greater attention from the scientific community. The need to reconsider botanical illustration through digital methods of graphic reproduction has been identified. Botanical illustration, as a specific genre, has developed historically into an important form of graphic reproduction. Its traditional and digital varieties, as well as botanical motifs, are effectively used by artists and designers in Ukraine and worldwide. At the current stage of digital media development, the language of graphic design is undergoing transformation, which influences both the approaches and stylistic features of botanical illustration. Through these changes, it is acquiring new functions and is being systematically integrated into design projects as an effective means of visual communication. Professional illustrators regularly employ botanical illustrations and motifs, combining traditional artistic techniques with digital technologies, mainly in print production. This fusion contributes to the stylistic enrichment of graphic design and opens new opportunities for creative realization. It is important to emphasize that digital botanical illustration, created using a stylus and tablet, represents a unique system of visual communication that significantly differs from traditional graphic techniques. This approach expands the possibilities for developing and applying botanical illustration, opening new perspectives for its advancement.
This study introduces LLM4DESIGN, a highly automated system for generating architectural and environmental design proposals. LLM4DESIGN, relying solely on site conditions and design requirements, employs Multi-Agent systems to foster creativity, Retrieval Augmented Generation (RAG) to ground designs in realism, and Visual Language Models (VLM) to synchronize all information. This system resulting in coherent, multi-illustrated, and multi-textual design schemes. The system meets the dual needs of narrative storytelling and objective drawing presentation in generating architectural and environmental design proposals. Extensive comparative and ablation experiments confirm the innovativeness of LLM4DESIGN's narrative and the grounded applicability of its plans, demonstrating its superior performance in the field of urban renewal design. Lastly, we have created the first cross-modal design scheme dataset covering architecture, landscape, interior, and urban design, providing rich resources for future research.
Patrizio Angelini, Therese Biedl, Markus Chimani
et al.
Not every directed acyclic graph (DAG) whose underlying undirected graph is planar admits an upward planar drawing. We are interested in pushing the notion of upward drawings beyond planarity by considering upward $k$-planar drawings of DAGs in which the edges are monotonically increasing in a common direction and every edge is crossed at most $k$ times for some integer $k \ge 1$. We show that the number of crossings per edge in a monotone drawing is in general unbounded for the class of bipartite outerplanar, cubic, or bounded pathwidth DAGs. However, it is at most two for outerpaths and it is at most quadratic in the bandwidth in general. From the computational point of view, we prove that testing upward-$k$-planarity is NP-complete already for $k=1$ and even for restricted instances for which upward planarity testing is polynomial. On the positive side, we can decide in linear time whether a single-source DAG admits an upward 1-planar drawing in which all vertices are incident to the outer face.
The draw of some knockout tournaments requires finding a perfect matching in a balanced bipartite graph. The problem becomes challenging with draw constraints: the two draw procedures used in sports are known to be non-uniformly distributed (the feasible matchings are not equally likely), which may threaten fairness. We compare the biases of both mechanisms, each of them having two forms, for reasonable subsets of balanced bipartite graphs up to 16 nodes. An optimal mechanism exist in the draw of quarterfinals under reasonable restrictions. The UEFA Champions League Round of 16 draw is verified to apply the least distorted design among the four available options between the 2003/04 and 2023/24 seasons. However, a considerable scope remains to improve these randomisation procedures.
Closed-loop optimal control design for high-dimensional nonlinear systems has been a long-standing challenge. Traditional methods, such as solving the associated Hamilton-Jacobi-Bellman equation, suffer from the curse of dimensionality. Recent literature proposed a new promising approach based on supervised learning, by leveraging powerful open-loop optimal control solvers to generate training data and neural networks as efficient high-dimensional function approximators to fit the closed-loop optimal control. This approach successfully handles certain high-dimensional optimal control problems but still performs poorly on more challenging problems. One of the crucial reasons for the failure is the so-called distribution mismatch phenomenon brought by the controlled dynamics. In this paper, we investigate this phenomenon and propose the Progressive Optimal Path Sampling (POPS) method to mitigate this problem. We theoretically prove that this enhanced sampling strategy outperforms both the vanilla approach and the widely used Dataset Aggregation (DAgger) method on the classical linear-quadratic regulator by a factor proportional to the total time duration. We further numerically demonstrate that the proposed sampling strategy significantly improves the performance on tested control problems, including the optimal landing problem of a quadrotor and the optimal reaching problem of a 7 DoF manipulator.
The Heating Neutral Beam Injectors (HNBs) for ITER will have to deliver 16.7 MW beams of H/D particles at 1 MeV energy. The beams will be produced from H-/D- ions, generated by a radiofrequency plasma source coupled to an ion acceleration system. A prototype of the ITER HNB ion source is being tested in the SPIDER experiment, part of the ITER Neutral Beam Test Facility at Consorzio RFX. Reaching the design targets for beam current density and fraction of coextracted electrons is only possible by evaporating cesium in the source, in particular on the plasma facing grid (PG) of the acceleration system. In this way the work function of the surfaces decreases, significantly increasing the amount of surface reactions that convert neutrals and positive ions into H-/D-. It is then of paramount importance to monitor the density of negative ions and the density of Cs in the proximity of the PG. Monitoring the Cs spatial distribution along the PG is also essential to guarantee the uniformity of the beam current. In SPIDER, this is possible thanks to the Cavity Ringdown Spectroscopy (CRDS) and the Laser absorption Spectroscopy diagnostics (LAS), which provide line-integrated measurements of negative ion density and neutral, ground state Cs density, respectively. The paper discusses the CRDS and LAS measurements as a function of input power and of the magnetic and electric fields used to reduce the coextraction of electrons. Negative ion density data are in qualitative agreement with the analogous measurements in Cs-free conditions. In agreement with simulations, Cs density is peaked in the center of the source; a top/bottom non uniformity is also present. Several effects of plasma on Cs deposition and negative ion production are presented.
We examine the role of textual data as study units when conducting causal inference by drawing parallels between human subjects and organized texts. %in human population research. We elaborate on key causal concepts and principles, and expose some ambiguity and sometimes fallacies. To facilitate better framing a causal query, we discuss two strategies: (i) shifting from immutable traits to perceptions of them, and (ii) shifting from some abstract concept/property to its constituent parts, i.e., adopting a constructivist perspective of an abstract concept. We hope this article would raise the awareness of the importance of articulating and clarifying fundamental concepts before delving into developing methodologies when drawing causal inference using textual data.
<span id="docs-internal-guid-a2bb7164-7fff-f94a-ec65-b31dc844c30a"><span>La producción de exposiciones temporales o itinerantes consume una gran cantidad de recursos materiales, la mayoría de difícil o imposible reciclado. En la metodología de trabajo de este sector, resulta clave incorporar la sostenibilidad y los conceptos provenientes del diseño sostenible y la economía circular para analizar y prever el impacto de los materiales en todas las fases del proceso de conceptualización y diseño. La reducción y reutilización de soportes, así como la sustitución de los clásicos materiales plásticos por otros de producción ecológica, reutilizables y reciclables, como la madera, el cartón y el papel de procedencia sostenible, son excelentes alternativas a los soportes habituales. La innovación en mejorar los materiales plásticos para ofrecer alternativas de menor impacto y las aplicaciones tecnológicas también ofrecen alternativas para limitar el uso de soportes de comunicación efímeros.</span></span>
Oswin Aichholzer, Matias Korman, Yoshio Okamoto
et al.
The total angular resolution of a straight-line drawing is the minimum angle between two edges of the drawing. It combines two properties contributing to the readability of a drawing: the angular resolution, which is the minimum angle between incident edges, and the crossing resolution, which is the minimum angle between crossing edges. We consider the total angular resolution of a graph, which is the maximum total angular resolution of a straight-line drawing of this graph. We prove that, up to a finite number of well specified exceptions of constant size, the number of edges of a graph with $n$ vertices and a total angular resolution greater than $60^{\circ}$ is bounded by $2n-6$. This bound is tight. In addition, we show that deciding whether a graph has total angular resolution at least $60^{\circ}$ is NP-hard.
We study planar straight-line drawings of graphs that minimize the ratio between the length of the longest and the shortest edge. We answer a question of Lazard et al. [Theor. Comput. Sci. 770 (2019), 88--94] and, for any given constant $r$, we provide a $2$-tree which does not admit a planar straight-line drawing with a ratio bounded by $r$. When the ratio is restricted to adjacent edges only, we prove that any $2$-tree admits a planar straight-line drawing whose edge-length ratio is at most $4 + \varepsilon$ for any arbitrarily small $\varepsilon > 0$, hence the upper bound on the local edge-length ratio of partial $2$-trees is $4$.
Based on the Chinese historical sunspots drawings, a data set consisting of the scanned images and all their digitized parameters from 1925 to 2015 have been constructed. In this paper, we briefly describe the developmental history of sunspots drawings in China. This paper describes the preliminary processing processes that strat from the initial data (inputing to the scanning equipment) to the parameters extraction, and finally summarizes the general features of this dataset. It is the first systematic project in Chinese solar-physics community that the historical observation of sunspots drawings were digitized. Our data set fills in an almost ninety years historical gap, which span 60 degrees from east to west and 50 degrees from north to south and have no continuous and detailed digital sunspot observation information. As a complementary to other sunspots observation in the world, our dataset provided abundant information to the long term solar cycles solar activity research.
The topic of this special issue of FormAkademisk is drawing. While the issue is hosted by the Oslo National Academy of the Arts (KHiO), the contributors have come in from different urban locations in Norway, including Volda, Trondheim, Bergen and Oslo. We would like to use this occasion to extend our thanks to the external peer-reviewers. They have helped in bringing the issue to its present level of quality
We present sketch-rnn, a recurrent neural network (RNN) able to construct stroke-based drawings of common objects. The model is trained on thousands of crude human-drawn images representing hundreds of classes. We outline a framework for conditional and unconditional sketch generation, and describe new robust training methods for generating coherent sketch drawings in a vector format.
Dan Hefetz, Christopher Kusch, Lothar Narins
et al.
We consider the strong Ramsey-type game $\mathcal{R}^{(k)}(\mathcal{H}, \aleph_0)$, played on the edge set of the infinite complete $k$-uniform hypergraph $K^k_{\mathbb{N}}$. Two players, called FP (the first player) and SP (the second player), take turns claiming edges of $K^k_{\mathbb{N}}$ with the goal of building a copy of some finite predetermined $k$-uniform hypergraph $\mathcal{H}$. The first player to build a copy of $\mathcal{H}$ wins. If no player has a strategy to ensure his win in finitely many moves, then the game is declared a draw. In this paper, we construct a $5$-uniform hypergraph $\mathcal{H}$ such that $\mathcal{R}^{(5)}(\mathcal{H}, \aleph_0)$ is a draw. This is in stark contrast to the corresponding finite game $\mathcal{R}^{(5)}(\mathcal{H}, n)$, played on the edge set of $K^5_n$. Indeed, using a classical game-theoretic argument known as \emph{strategy stealing} and a Ramsey-type argument, one can show that for every $k$-uniform hypergraph $\mathcal{G}$, there exists an integer $n_0$ such that FP has a winning strategy for $\mathcal{R}^{(k)}(\mathcal{G}, n)$ for every $n \geq n_0$.
We prove that for every graph $G$, given fixed locations for the vertices of $G$ in $\mathbb{Z}^3$, there is a three-dimensional grid-drawing of $G$ with one bend per edge. The best previous bound was three bends per edge.
René van Bevern, Iyad Kanj, Christian Komusiewicz
et al.
Given a hypergraph $H$, the Planar Support problem asks whether there is a planar graph $G$ on the same vertex set as $H$ such that each hyperedge induces a connected subgraph of $G$. Planar Support is motivated by applications in graph drawing and data visualization. We show that Planar Support is fixed-parameter tractable when parameterized by the number of hyperedges in the input hypergraph and the outerplanarity number of the sought planar graph. To this end, we develop novel structural results for $r$-outerplanar triangulated disks, showing that they admit sequences of separators with structural properties enabling data reduction. This allows us to obtain a problem kernel for Planar Support, thus showing its fixed-parameter tractability.