Strokes2Surface: Recovering Curve Networks From 4D Architectural Design Sketches

TL;DR

Strokes2Surface addresses the challenge of turning imprecise 4D architectural sketches into usable geometry for digital modeling by introducing an offline pipeline that first classifies strokes as Shape or Scribble and then clusters them to form a well-connected curve network and corresponding surface patches. The method relies on a two-path clustering approach, topology recovery, and cycle-aware surfacing guided by Scribble clusters, all driven by a 4D sketching interface that records rich metadata. Key contributions include the Shape/Scribble dichotomy with hand-engineered features, a robust 3D sketch consolidation method, a new dataset of 4D architectural sketches, and comprehensive ablation and user studies showing usability and performance advantages over prior art. The approach enables efficient, offline reconstruction that preserves design intent and yields patch-based surfaces suitable for BIM workflows, marking a significant step toward bridging concept design and digital modeling in architecture.

Abstract

We present Strokes2Surface, an offline geometry reconstruction pipeline that recovers well-connected curve networks from imprecise 4D sketches to bridge concept design and digital modeling stages in architectural design. The input to our pipeline consists of 3D strokes' polyline vertices and their timestamps as the 4th dimension, along with additional metadata recorded throughout sketching. Inspired by architectural sketching practices, our pipeline combines a classifier and two clustering models to achieve its goal. First, with a set of extracted hand-engineered features from the sketch, the classifier recognizes the type of individual strokes between those depicting boundaries (Shape strokes) and those depicting enclosed areas (Scribble strokes). Next, the two clustering models parse strokes of each type into distinct groups, each representing an individual edge or face of the intended architectural object. Curve networks are then formed through topology recovery of consolidated Shape clusters and surfaced using Scribble clusters guiding the cycle discovery. Our evaluation is threefold: We confirm the usability of the Strokes2Surface pipeline in architectural design use cases via a user study, we validate our choice of features via statistical analysis and ablation studies on our collected dataset, and we compare our outputs against a range of reconstructions computed using alternative methods.

Figures (14)

• Figure 1: 3D stroke formation through ray casting. As 2D strokes are drawn on the tablet's surface, rays originating from the camera's viewpoint are intersected the selected canvas, and the resulting 3D point lying on the canvas is stored as the continuation of the corresponding 3D stroke polyline vertices.
• Figure 2: An example of a conventional 2D drawing of an architectural object portraying two types of strokes as the design: strokes outlining the boundaries and those marking the enclosed areas. Images' source: iStock. Credit: SireAnko. Licenses purchased by the first author.
• Figure 3: The top-left image represents the original sketch of a curved wall, created using the 4D drawing interface for theater stage design purposes, and the rest are the visualizations of raw, scalar-valued features of the sketch strokes that are color-coded from pink to green using a colormap that maps feature values into distinct colors according to their magnitude. The gradient ranging from pink to green denotes the increase in feature values, with pink representing lower values and green representing higher values. Below each image, $\min$ and $\max$ of each feature value are noted. Input sketch: Ingrid Erb.
• Figure 4: Left: the original sketch if a pavilion with three legs. Right: the classifier's predictions on strokes with Shape Strokes colored in blue and Scribble strokes in red.
• Figure 5: (a) A sample sketch of a dome-like architectural object created by participant P01. (b) The outcome of the classification process, where strokes predicted as Scribble are highlighted in red, and those identified as Shape are marked in blue. (c) The sketch after removal of the Scribble strokes, leaving only the Shape strokes in their original color. (d) The result of clustering the Shape strokes into distinctive groups, each representing a boundary or edge, with each group displayed in a unique color for clarity.