iCanDraw? Using Sketch Recognition and Corrective Feedback to Assist a User in Drawing Human Faces

Bibliography Information:

Dixon, Prasad, Hammond. "iCanDraw? Using Sketch Recognition and Corrective Feedback to Assist a User in Drawing Human Faces." 2010. CHI Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. Pages 897-906. ACM New York, NY, USA ©2010

Link to Article: http://dl.acm.org/citation.cfm?id=1753459

iCanDraw? is a system that uses the principles of sketch recognition to teach participants about basic techniques behind sketching portraits. Researchers use a facial recognition library that provides them with 40 data points including contours of the mouth, wings of the nose, contour of each eyebrow, etc. Researchers then make manual corrections if necessary, and these points are then derived again from a user's input sketch based on the same image and uses bounding boxes and line end points to derive the "correctness" of the sketch. Correctness, in this instance, is defined by a user's similarity to the presented portrait in basic contour composition of major facial features like eyes, mouth, and nose. Five users participated in this study who identified themselves as unable to draw well, and they recognized the program's ability to discern correctness of their sketch's composition. They also appreciated contextual feedback based on areas of improvement in their sketch.

I think this paper shows a creative and productive use of facial recognition algorithms beyond the obvious identification of individuals. I appreciated the inclusion of the drawing instructor's evaluation of this program and his thoughts on its approach of teaching. Some of the methods of evaluation could have better been discussed with a little more detail, such as how they perceived the system's ability to improve the participants' skill. Additionally, I found it interesting that several of the participants' initial freehand drawing showed focus on the subject's clothing. In the case of the baby, even individual lines were drawn to indicate the grid-like pattern of the baby's shirt. However, in the second sketch shown after the participant has used iCanDraw?'s systems we see considerably less focus on non-facial features, presumably due to the program's focus in facial features. This could be used in an educational setting where these algorithms can be applied to areas where the participant is known to have skill gaps, since this focus clearly shows that users are encouraged to place more attention in some areas over others. Further evaluations of shifts in focus as a result of using this system could yield interesting results.