1. CircuitBoard Shane Zamora Eyrún A. Eyjólfsdóttir University of California, Santa Barbara Department of Computer Science Sketch-Based Logic Circuit Design

2. Introduction Recognition System UI Design Demo! Outline

3. Digital circuit CAD applications are too complex. Students and professionals prefer working on a pad of paper or a whiteboard. However, CAD applications can provide valuable tools for design work: Debugging!

4. Better idea: Use computers to design computers. Tablet PCs can not only provide the same feel of using a pad of paper, but they can compute. Provide a CAD application for digital circuit design that augments a clean, paper- like interface with real-time simulation and feedback. Our Goal:

5. Previous Work Surprisingly, only a few programs have been developed specifically for sketch-based digital logic design. However, many papers that introduce new algorithms or frameworks do mention digital logic circuits as potential applications. E-Chalk Plugin - Liwicki, M. and Knipping, L. 2005 Labeler - Alvarado, C. 2007 Two best applications ( that we could find ):

6. Previous Work ( Continued ) E-Chalk Plugin Our Goal: Augments a clean, paper- like interface with real-time simulation and feedback. Images from Liwikci, M. and Knipping, Recognizing and Simulating Sketched Logic Circuits. Proc. KES 2005, 588-594 Labeler Alvarado, 2007 Our Goal: Augments a clean, paper- like interface with real-time simulation and feedback. Alvarado C. Sketch Recognition for Digitial Circuit Design in the Classroom. Proc. Invited Workshop on Pen-Centric Computing Research, 2007.

7. CircuitBoard Recognizes all types of logic gates. Flexible input system does not constrain the user. Circuit simulation / debugging. Modeless input, modal visualization.

8. Introduction Recognition System UI Design Demo! Outline

9. Simulation Mode Normal Mode Gate Mode Visualization Modes Normal Mode provides a clean, unobstructed surface for logic design. Simulation mode allows the user to interact with and debug their work. Gate mode allows the user the review recognition accuracy and signal flow.

10. UI Controls Gate ModeNormal ModeSimulation Mode CtrlShift Buttons Caps Lock Label fades in and out quickly – tap to change gate to next most likely interpretation

11. Introduction Recognition System UI Design Demo! Outline

12. Step 1: Stroke Segmentation Raw stroke data is not very useful, so we segment it into lines and curves. We use the scale-space feature point detection method, by Sezgin and Davis from MIT. (2004) Q: How well did this work? A: Not as well as we'd like.

13. Images/Data from Sezgin, T. M. and Davis, R. Scale-space based feature point detection for digital ink. Proc. Making Pen-Based Interaction Intelligent and Natural 2004. Gaussian Strength (σ) Feature Point Count Raw Stroke Segmented Stroke

15. Ideally fragmentation would give us

16. In reality we get Our solution: Use a Computer Vision technique instead of relational constraints.

18. Classification Training sample Query sample ● Invariant to: ✔ rotation ✔ scale ✔ small distortion

19. Introduction Recognition System UI Design Demo! Outline

20. Thank You! Questions? Contact Info Shane Zamora char42@cs.ucsb.edu Eyrún A. Eyjólfsdóttir eyrun@cs.ucsb.edu Project Website http://cs.ucsb.edu/~arch/sketch/circuitboard/