Search arXiv⌕ Search
Home/Archive/arXiv

arXiv · 2609.30955

Flow-TAG: Flow-based conditional latent transport for accurate spline approximation and data compression

Roman Pavelkin·Luis A. Zavala-Mondragon·Fons van der Sommen
View original source ↗Read source document ↗Export metadata ↓

Abstract

Robust curve fitting is essential in computer-aided design for transforming noisy, discrete data into accurate geometric models that ensure numerical stability across engineering workflows. B-spline models have become the industry standard for this task, offering a flexible and reliable framework characterized by local control and smooth shape representation. This paper presents flow-TAG--a data-driven framework based on a generative flow model with a 1D U-Net backbone capable of mapping the geometry of a curve to the optimal parametrization for cubic B-splines. By leveraging learned geometric patterns, flow-TAG exhibits superior parameterization performance, robustness to noise in the input data, and strong generalization capability to previously unseen 2D and 3D curves drawn from distinct data distributions. Flow-TAG yields fitted curves that achieve the lower root-mean-square error (55% lower on average) and Hausdorff distance (52% lower on average) relative to state-of-the-art data-driven methods. In addition, we investigate the practical applicability of our generative framework in the compression of ECG signals for wearable devices. The proposed compression setup provides a compression ratio of 13 with the signal distortion of around 5%, which is acceptable in the field.

Explore related subjects

cs.CG

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Roman Pavelkin, Luis A. Zavala-Mondragon, Fons van der Sommen. 2026-09-25. Flow-TAG: Flow-based conditional latent transport for accurate spline approximation and data compression. https://arxiv.org/abs/2609.30955

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Geometric Optimization Parameterized by Piercing Complexity

Packing and covering problems for geometric regions have been studied under many notions of complexity, including VC-dimension, union complexity, shallow-cell complexity, and fatness. Although these restrictions often yield constant-factor approximation algorithms, they do not by themselves generally lead to PTASs. A recurring feature of known hardness constructions is that one region may be \emph{pierced} by many others: a region $B$ pierces $A$ when $A\setminus B$ is disconnected. We study geometric instances through the \emph{piercing degree}. Since piercing is symmetric for Jordan regions, this is the maximum degree of the corresponding piercing graph. Our main result is that, for every fixed piercing degree, the standard local-search algorithms give PTASs for the unweighted \emph{Discrete Independent Set} and \emph{Set Cover} problems. The proof constructs a sublinear balanced separator for an appropriate locality graph and applies it adaptively throughout the recursive local-search analysis. This guarantee depends only on the piercing degree; in particular, it places no bound on the number of components created by an individual piercing pair. We also prove a polynomial shallow-trace bound depending only on the piercing degree. As consequences, for every fixed piercing degree, weighted Set Cover admits a deterministic $C_r$-approximation and weighted Discrete Independent Set admits a deterministic $O(r+1)$-approximation. These results extend the known guarantees for non-piercing families and apply, for example, to axis-parallel rectangles when every rectangle is pierced by only a bounded number of other rectangles.

cs.CG↗
arXiv2026

Largest-Area Convex Quadrilateral in a $1.5$D Terrain

A $1.5$D terrain is a simple polygon bounded by a horizontal base and an $x$-monotone upper chain. We study the problem of finding a largest-area convex quadrilateral contained in an $n$-vertex terrain. We maximize area over the closure of the feasible nondegenerate quadrilaterals, allowing a triangular boundary optimum when necessary. Assuming that no three terrain vertices are collinear, we give a deterministic exact algorithm running in $O(n^2\log n)$ time and using $O(n)$ working space in the algebraic real-RAM. Among all optimum solutions, the algorithm returns a nondegenerate quadrilateral whenever one exists; otherwise, it returns a maximum-area terrain triangle. We also prove that a maximum-area axis-parallel rectangle contained in the terrain yields a tight $\frac12$-approximation and can be computed in $O(n\log n)$ time.

cs.CG↗
arXiv2026

Depth-Based Local Recentering of Maxmin Landmarks for Lazy Witness Persistence

We study a local post-processing rule for maxmin (farthest-point) landmark selection in lazy witness persistence. Maxmin landmarks induce nearest-landmark cells. Within each cell, the landmark is moved toward a data point of maximum empirical halfspace depth, with smaller moves in sparsely populated cells, and is then projected back to a data point in the same cell. The resulting support-weighted rule preserves one landmark per original maxmin cell. We prove a convex-core guarantee for the depth target and explicit worst-case cover bounds for the recentered landmarks. Experiments on planar synthetic loops and 120 MPEG-7 (Moving Picture Experts Group 7) silhouettes separate average landmark fit from extremal coverage. On the synthetic benchmark, recentering reduces mean nearest-landmark distance in all 900 matched trials; after aggregation over landmark budgets, the mean change is $-0.0138$ across 300 dataset/noise/seed blocks, while maximum cover increases by $0.0067$ on average and a thresholded first homology ($H_1$) count is unchanged. On MPEG-7, the image-level mean distance decreases for all 120 silhouettes, while maximum cover increases for 118. Thus the correction consistently improves average landmark representativeness but does not improve worst-case coverage; on the controlled synthetic benchmark, this trade-off occurs without a detectable change in the coarse $H_1$ summary.

cs.CG↗