[Paper Review] A Method for Visuo-Spatial Classification of Freehand Shapes Freely Sketched

Ney Renau-Ferrer, Céline Rémi|arXiv (Cornell University)|May 7, 2013

3D Surveying and Cultural Heritage3 references4 citations

TL;DR

This paper presents a novel method for visuo-spatial classification of freehand sketches by analyzing geometric and spatial properties of online-drawn shapes. Using a combination of shape normalization, feature extraction, and dissimilarity measurement against ideal geometric templates, the approach achieves over 95% accuracy in classifying sketches to reference shapes, demonstrating strong potential for multi-level sketch analysis in human-computer interaction and pattern recognition systems.

ABSTRACT

We present the principle and the main steps of a new method for the visuo-spatial analysis of geometrical sketches recorded online. Visuo-spatial analysis is a necessary step for multi-level analysis. Multi-level analysis simultaneously allows classification, comparison or clustering of the constituent parts of a pattern according to their visuo-spatial properties, their procedural strategies, their structural or temporal parameters, or any combination of two or more of those parameters. The first results provided by this method concern the comparison of sketches to some perfect patterns of simple geometrical figures and the measure of dissimilarity between real sketches. The mean rates of good decision higher than 95% obtained are promising in both cases.

Motivation & Objective

• To enable multi-level analysis of freehand sketches by integrating visuo-spatial, procedural, structural, and temporal parameters.
• To address the challenge of classifying irregular, user-drawn shapes that deviate from perfect geometric forms.
• To develop a robust framework for measuring dissimilarity between real sketches and ideal geometric patterns.
• To support applications in sketch-based user interfaces, educational tools, and pattern recognition by enabling accurate classification of freehand drawings.

Proposed method

• The method begins with online recording of freehand sketches to preserve spatial and temporal trajectory data.
• It normalizes the sketch by aligning its principal axis and scaling it to a standard size to reduce orientation and scale variability.
• Geometric features such as curvature, angularity, and contour continuity are extracted from the normalized sketch.
• A dissimilarity measure is computed between the normalized sketch and predefined ideal geometric templates (e.g., circles, squares).
• The classification decision is made by selecting the ideal shape with the minimum dissimilarity score.
• The approach supports multi-level analysis by combining visuo-spatial features with procedural and structural parameters.

Experimental results

Research questions

• RQ1 How can freehand sketches be effectively classified based on their visuo-spatial properties rather than just shape outline?
• RQ2 What level of accuracy can be achieved in matching irregular, user-drawn sketches to ideal geometric templates?
• RQ3 How can dissimilarity between real sketches and perfect geometric forms be quantitatively measured in a way that reflects human perception?
• RQ4 Can the method support multi-level analysis by integrating visuo-spatial, procedural, and structural parameters?
• RQ5 What is the performance of the method in distinguishing between different types of freehand-drawn shapes?

Key findings

• The method achieves mean classification accuracy higher than 95% when comparing freehand sketches to ideal geometric templates.
• The dissimilarity measure effectively captures perceptual differences between real sketches and perfect geometric forms.
• The normalization process significantly improves classification robustness by reducing sensitivity to scale and orientation.
• The approach demonstrates strong potential for integration into sketch-based interfaces and educational applications.
• The results confirm the feasibility of using visuo-spatial features as a foundation for multi-level sketch analysis.