Download presentation
Presentation is loading. Please wait.
Published byBranden Watkins Modified over 8 years ago
1
Resolving Ambiguities to Create a Natural Sketching Environment Christine Alvarado and Randall Davis MIT AI Laboratory
2
Christine Alvarado Our Model The Designer Sketches with Pen and Paper The Observer Interprets the Sketch The Observer and Designer Interact
3
Christine Alvarado Sketch Interpretation
4
Christine Alvarado Accuracy vs. Freedom Drawing Freedom Recognition Difficulty Menu “Put That There” Single Stroke Recognition Free Sketch ASSIST
5
Christine Alvarado Accuracy and Freedom Smarter interpretation increases accuracy Better interaction design increases perceived freedom
6
Christine Alvarado Resolving Ambiguities Levels of Interpretation Fluid Interpretation Commitment to an Interpretation
7
Christine Alvarado 3 Stages of Interpretation Recognition Reasoning Resolution
8
Christine Alvarado Recognition Generate All Possible Interpretations: –Circle –Circular Body –Pin Joint
9
Christine Alvarado Reasoning: Heuristics Temporal Evidence Simpler Is Better Context Domain Knowledge User Feedback
10
Christine Alvarado Reasoning: Heuristics Temporal Evidence Simpler Is Better Context Domain Knowledge User Feedback
11
Christine Alvarado Reasoning: Heuristics Temporal Evidence Simpler Is Better Context Domain Knowledge User Feedback 1 arrow or 3 rods?
12
Christine Alvarado Reasoning: Heuristics Temporal Evidence Simpler Is Better Context Domain Knowledge User Feedback
13
Christine Alvarado Reasoning: Heuristics Temporal Evidence Simpler Is Better Context Domain Knowledge User Feedback
14
Christine Alvarado Reasoning: Heuristics Temporal Evidence Simpler Is Better Context Domain Knowledge User Feedback
15
Christine Alvarado Reasoning: Heuristics Temporal Evidence Simpler Is Better Context Domain Knowledge User Feedback Total Score
16
Christine Alvarado Resolution
17
Christine Alvarado Resolution 0 3 6
18
Christine Alvarado Resolution 0 3 6
19
Christine Alvarado Resolution 0 0 55 10
20
Christine Alvarado Resolution 0 5 10 0 5
21
Christine Alvarado Resolution 0 000 3 10 3 6 3 6 3 6 6
22
Christine Alvarado Resolution 0 000 3 10 3 6 3 6 3 6 6
23
Christine Alvarado Resolution 0 000 3 10 3 6 3 6 3 6 0 3 6 6 8
24
Christine Alvarado Resolution 0 000 3 10 3 6 3 6 3 6 0 3 6 6 8
25
Christine Alvarado Resolution 0 000 3 10 3 6 3 6 3 6 0 3 6 6 8
26
Christine Alvarado Relies heavily on bottom-up recognition Limitations line ??? Line + Line + Line + ??? ???
27
Christine Alvarado Relies heavily on bottom-up recognition Heuristics all weighted equally Limitations H1: H2:Prefer objects drawn with contiguous strokes Prefer interpretations resulting in fewer objects
28
Christine Alvarado Relies heavily on bottom-up recognition Heuristics all weighted equally Limitations H1: H2:Prefer objects drawn with contiguous strokes Prefer interpretations resulting in fewer objects
29
Christine Alvarado Relies heavily on bottom-up recognition Heuristics all weighted equally Limitations H1: H2:Prefer objects drawn with contiguous strokes Prefer interpretations resulting in fewer objects
30
Christine Alvarado Structured Application of Context Blackboard recognition architecture Heuristics applied probabilistically
31
Christine Alvarado Recognition Blackboard Sketch Blackboard Line(l1)Line(l2)Line(l3) Line(l5)Line(l4)Line(l7) Connects(l1, l2) Arrow(a1) Force(f1) Stroke(s2)Stroke(s1)Stroke(s3)Stroke(s5)Stroke(s7)Stroke(s6)Stroke(s4) Connects(l4, l5) Connects(l7, l4) Forces push bodies Line(l6) Connects(l5, l6) Connects(l6, l7) Polygon(p1) Body(b1)
32
Christine Alvarado Bayesian Network Structure Heuristics influence prior Line1Line2Line3 Property1Property2 … … Low-level information influences recognition
33
Christine Alvarado Related Work Gross and Do (1996) Landay and Meyers (2001) Stahovich (1998) Matsakis (1999)
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.