Stroke of Surprise: Progressive Semantic Illusions in Vector Sketching
2026-02-12 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors created a new kind of drawing trick called Progressive Semantic Illusions, where one sketch changes meaning as more lines are added. They developed a system named Stroke of Surprise that carefully adjusts the drawing so it looks like one object at first (like a duck) and then transforms into another (like a sheep) as more strokes appear. Their method cleverly balances both images by tweaking the early lines to fit both ideas, using a special optimization technique and a new way to keep the shapes from covering each other improperly. Tests show their approach works better than previous methods at making these clever visual transformations over time.
vector sketchingsemantic transformationvisual illusionsScore Distillation Sampling (SDS)dual-constraint optimizationsemantic prefix strokesOverlay Lossjoint optimizationvisual anagramstemporal dimension in drawing
Authors
Huai-Hsun Cheng, Siang-Ling Zhang, Yu-Lun Liu
Abstract
Visual illusions traditionally rely on spatial manipulations such as multi-view consistency. In this work, we introduce Progressive Semantic Illusions, a novel vector sketching task where a single sketch undergoes a dramatic semantic transformation through the sequential addition of strokes. We present Stroke of Surprise, a generative framework that optimizes vector strokes to satisfy distinct semantic interpretations at different drawing stages. The core challenge lies in the "dual-constraint": initial prefix strokes must form a coherent object (e.g., a duck) while simultaneously serving as the structural foundation for a second concept (e.g., a sheep) upon adding delta strokes. To address this, we propose a sequence-aware joint optimization framework driven by a dual-branch Score Distillation Sampling (SDS) mechanism. Unlike sequential approaches that freeze the initial state, our method dynamically adjusts prefix strokes to discover a "common structural subspace" valid for both targets. Furthermore, we introduce a novel Overlay Loss that enforces spatial complementarity, ensuring structural integration rather than occlusion. Extensive experiments demonstrate that our method significantly outperforms state-of-the-art baselines in recognizability and illusion strength, successfully expanding visual anagrams from the spatial to the temporal dimension. Project page: https://stroke-of-surprise.github.io/