Computer Vision Readings Discussion and Activity Susie Wise 11/11/2004
Computer Vision for Interactive Computer Graphics From multimodal to looking specifically at methods for using VISION Vision can be a powerful interface device because of its potential for sensing Body position Head orientation Direction of gaze Pointing commands Gesture
Computer Vision for Interactive Computer Graphics (2) They make the connection to interactive graphics as arena in which to work Successful because limited context (limits recognition req’d) human is in the loop (can make adjustments because they are getting graphical feedback)
Computer Vision for Interactive Computer Graphics (3) Areas of exploration Large object tracking Shape recognition Motion analysis Small object tracking
A Design Tool for Camera-based Interaction Recognizing that image classifiers are hard to design and develop, they have a made tool Goal for ordinary interface designers to work with camera-based interaction Introduce “crayons” which allow UI designer to generate training data quickly by “painting” classification over sample images
Design Principles Use context to limit the range of visual interpretations Take advantage of graphical feedback (to make adjustments, constrain input) Simplify image classifier
Activity Divide into small groups Pick an arena where vision-tracking could be fruitful (based on one in the articles or come up with a realm of your own) Brainstorm application scenarios As with Milestone #1 ideation, go for breadth Report back to full group Range of ideas Consensus on best to pursue