Reading 3

Citation:

Dixon, Daniel, Manoj Prasad, and Tracy Hammond. "iCanDraw: using sketch recognition and corrective feedback to assist a user in drawing human faces. “Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. ACM, 2010.

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

Summary:

The first system for using computer aided instruction for assisting a student to draw human faces is presented. It uses face and sketch recognition concepts. The system provides a step by step instruction and generated feedback. While L-mode , R-mode is more equipped to understand visual data. Each step provides a written explanation describing current task.

The first iteration had prep-sketching, Embossing strokes and a separate drawing area. Second iteration is improved through nine design principles. It has the options like: Check my work, Show suggestions, Next, aspect ratio checking and corrective feedback. Each stroke is re-sampled to make the data points equidistant from one another along the stroke. Whether or not correction is needed is judged by whether the windows distance is greater than one standard deviation from the average. In quantitative evaluation, all participants found step-by-step approach useful.

Discussion:

I found the nine design principals very thought provoking. Supplying feedback only when asked for is a great idea as it can disturb the thought process and ruin the whole drawing. I also feel that the sample spacing for people using the system and giving feedback should have been more in size to have better clarity of where we stand.

Reading 1

Citation:

Valentine, Stephanie, et al. "Mechanix: a sketch-based tutoring and grading system for free-body diagrams." AI Magazine 34.1 (2012): 55.

Publication Link: http://www.aaai.org/ojs/index.php/aimagazine/article/view/2437

Summary:

Mechanix uses Integrated scribble gesture for removing complete or a part of shape and primitive recognition as feedback method. Receiving subtle feedback leads students to draw correctly recognized trusses or bodies. For truss matching, graph data structure is used. For comparing, both shapes are reassembled to contain 64 evenly spaced points. A combination of the Hausdorff distance, modified Hausdorff distance and Tanimato coefficient determines  similarities between new sketch and posted solution of body comparison. Mechanix follows a distributed client server architecture. Data is transferred between client and server as XML over HTTP. 

Discussion:

Generating system of equations apart from just the sketch matching is a promising feature that mechanix has. Definitely majority students preferring to use Mechanix again is a big success. But it needs to be thought over as to why even a few students did not take it forward with Mechanix. The real time suggestion providing and interpretation of answers is commendable.

Reading 2

Citation:

 Martin, Field,  Valentine, Stephanie, Linsey, Julie, Hammond, Tracy. "Sketch recognition algorithms for comparing complex and unpredictable shapes." IJCAI. 2011.

Link: http://www.researchgate.net/profile/Tracy_Hammond/publication/220812144_Sketch_Recognition_Algorithms_for_Comparing_Complex_and_Unpredictable_Shapes/links/0deec529f76e510edf000000.pdf

Summary:

Mechanix works on both free body diagrams and Trusses. For non truss shapes, it uses a geometric approach. A combination of the Hausdorff distance, modified Hausdorff distance and Tanimato coefficient determines  similarities between new sketch and posted solution of body comparison. Adjacency graph is used for comparing the student's truss and reference solution, in addition to the sketch.

Discussion:

The paper efficiently describes identification and comparison of body and trusses. Using adjacency graph and BFS seem reasonable for comparison. This paper deals with free body diagrams and trusses only. The scope could be increased to other shapes as well.