Abstract

Sketch recognition systems are currently being developed for many domains, but can be time consuming to build if they are to handle the intricacies of each domain. In order to aid sketch-based user interface developers, we have developed tools to simplify the development of a new sketch recognition interface. We created LADDER, a language to describe how sketched diagrams in a domain are drawn, displayed, and edited. We then automatically transform LADDER structural descriptions into domain specific shape recognizers, editing recognizers, and shape exhibitors for use in conjunction with a domain independent sketch recognition system, creating a sketch recognition system for that domain. We have tested our framework by writing several domain descriptions and automatically generating a domain specific sketch recognition system from each description.

Keywords: Object recognition; Object modeling; Shape; Hierarchical scene analysis; Knowledge representation; Representation languages; Representations; Vision and scene understanding; User-centered design

Article Outline

1. Introduction

2. LADDER

2.1. Description limitations

2.2. Shape definition

2.2.1. Hierarchical shape definitions

2.2.2. Abstract shape definitions

2.2.3. Shape groups

2.3. Language contents

2.3.1. Predefined shapes

2.3.2. Predefined constraints

2.3.3. Predefined editing behaviors, actions, and triggers

2.3.4. Predefined display methods

2.4. Vectors

3. Multi-domain recognition system

3.1. Recognition of primitive shapes

3.2. Recognition of domain shapes

3.3. Editing recognition

3.4. Constraint solver

4. Code generation

5. Evaluation

6. Related work

6.1. Visual or sketching languages

6.2. Building recognition systems

6.3. Translation

7. Future work

8. Contributions

Acknowledgements

References