StripMaker: Perception-driven Learned Vector Sketch Consolidation

Abstract

Artist sketches often use multiple overdrawn strokes to depict a single intended curve. Humans effortlessly mentally consolidate such sketches by detecting groups of overdrawn strokes and replacing them with the corresponding intended curves. While this mental process is near instantaneous, manually annotating or retracing sketches to communicate this intended mental image is highly time consuming; yet most sketch applications are not designed to handle overdrawing and can only operate on overdrawing-free, consolidated sketches. We propose StripMaker , a new and robust learning based method for automatic consolidation of raw vector sketches. We avoid the need for an unsustainably large manually annotated learning corpus by leveraging observations about artist workflow and perceptual cues viewers employ when mentally consolidating sketches. We train two perception-aware classifiers that assess the likelihood that a pair of stroke groups jointly depicts the same intended curve: our first classifier is purely local and only accounts for the properties of the evaluated strokes; our second classifier incorporates global context and is designed to operate on approximately consolidated sketches. We embed these classifiers within a consolidation framework that leverages artist workflow: we first process strokes in the order they were drawn and use our local classifier to arrive at an approximate consolidation output, then use the contextual classifier to refine this output and finalize the consolidated result. We validate StripMaker by comparing its results to manual consolidation outputs and algorithmic alternatives. StripMaker achieves comparable performance to manual consolidation. In a comparative study participants preferred our results by a 53% margin over those of the closest algorithmic alternative (67% versus 14%, other/neither 19%).

Region‐Aware Simplification and Stylization of 3D Line Drawings

“…Unless additional post‐processing is applied, StripMaker [LABS23] produces outputs containing gaps between regions such that finding correspondences to the original regions is difficult. Compositing a stylized version of this line drawing with the original regions creates inconsistencies.…”

“…[ST90, LMLH07], but does not produce vector graphics suitable for shape‐based stylization or editing. Methods that aim to extract the perceived or intended curve network from human‐drawn sketches, either from input raster images (i.e., line drawing vectorization [FLB16,SBBB20]) or from input vector curves (i.e., stroke consolidation [BTS05, OK11, SSISI16, PvMLV*21, LABS23]) require some degree of line simplification due to the overdrawn nature of human sketches. However, simplification is limited to aggregating nearby lines, rather than simplifying overall structure and regions.…”

“…For much more complex drawings, such as the car drawing, preprocessing can take up to 10 minutes, and optimization up to 25 minutes. The main bottleneck during optimization is in rendering then differentiating through the density image, which requires ren- Figure 13: Unless additional post-processing is applied, StripMaker [LABS23] produces outputs containing gaps between regions such that finding correspondences to the original regions is difficult. Compositing a stylized version of this line drawing with the original regions creates inconsistencies.…”

“…StripMaker. StripMaker [LABS23] is a sketch consolidation method that replaces overdrawn strokes with a single curve corresponding to a perceived intended curve. We preprocess our input vector drawings as described in Section 3.5, i.e., we filter out short lines and small regions.…”

Strokes2Surface: Recovering Curve Networks From 4D Architectural Design Sketches

“…[LRS18] uses human perception principles and artistic practice observations for stroke clustering, employing angular and proximity scores with HDBSCAN for incremental merging, followed by local cluster refinement and curve fitting. Given a vector sketch with multiple overdrawn strokes, [LABS23] consolidates it through two classifiers: the first evaluates strokes locally, while the second, incorporates global context, refining the first classifier's consolidation output. The second step in our pipeline shares similarities with existing 2D sketch consolidation methods.…”