Paper
Wolin, Aaron, Brian Eoff, and Tracy Hammond. "ShortStraw: A Simple and Effective Corner Finder for Polylines." SBM. 2008.
Direct Link: http://www.researchgate.net/profile/Tracy_Hammond/publication/220772398_ShortStraw_A_Simple_and_Effective_Corner_Finder_for_Polylines/links/0deec529f76e58d523000000.pdf
Notes
This is a fun and interesting corner detection algorithm that's also very simple.
Wednesday, September 30, 2015
Monday, September 28, 2015
Reading 13: Domain-Independent Recognition
Paper
Yu, Bo, and Shijie Cai. "A domain-independent system for sketch recognition." Proceedings of the 1st international conference on Computer graphics and interactive techniques in Australasia and South East Asia. ACM, 2003.
Publication Link: http://dl.acm.org/citation.cfm?id=604499
Yu, Bo, and Shijie Cai. "A domain-independent system for sketch recognition." Proceedings of the 1st international conference on Computer graphics and interactive techniques in Australasia and South East Asia. ACM, 2003.
Publication Link: http://dl.acm.org/citation.cfm?id=604499
Friday, September 25, 2015
Reading 12: Sketch-Based Interfaces
Paper
Sezgin, Tevfik Metin, Thomas Stahovich, and Randall Davis. "Sketch based interfaces: early processing for sketch understanding." ACM SIGGRAPH 2006 Courses. ACM, 2006.
Publication Link: http://dl.acm.org/citation.cfm?id=1185783
Notes
This paper is a good example of how several of these sketch features and algorithms can be combined into a user interface.
Sezgin, Tevfik Metin, Thomas Stahovich, and Randall Davis. "Sketch based interfaces: early processing for sketch understanding." ACM SIGGRAPH 2006 Courses. ACM, 2006.
Publication Link: http://dl.acm.org/citation.cfm?id=1185783
Notes
This paper is a good example of how several of these sketch features and algorithms can be combined into a user interface.
Wednesday, September 23, 2015
Reading 11 - No Rubine? Geometric-Based Features
Paper
Paulson, Brandon, et al. "What!?! no Rubine features?: using geometric-based features to produce normalized confidence values for sketch recognition." HCC Workshop: Sketch Tools for Diagramming. 2008.
Direct Link: https://www.cs.auckland.ac.nz/research/conferences/skekchws/proceedings/vlhcc_stws_p57.pdf
Notes
Continuing our discussion of different sketch features, this paper looks at some geometric-based features.
Paulson, Brandon, et al. "What!?! no Rubine features?: using geometric-based features to produce normalized confidence values for sketch recognition." HCC Workshop: Sketch Tools for Diagramming. 2008.
Direct Link: https://www.cs.auckland.ac.nz/research/conferences/skekchws/proceedings/vlhcc_stws_p57.pdf
Notes
Continuing our discussion of different sketch features, this paper looks at some geometric-based features.
Monday, September 21, 2015
Reading 10 - Visual Similarity of Gestures
Paper
Long Jr, A. Chris, et al. "Visual similarity of pen gestures." Proceedings of the SIGCHI conference on Human Factors in Computing Systems. ACM, 2000.
Publication Link: http://dl.acm.org/citation.cfm?id=332458
Notes
These authors have a lot of work in the area of gesture recognition using different form factors and inputs. This paper discusses some of the background for their work, examining visual similarity of gestures and how this may be determined.
Long Jr, A. Chris, et al. "Visual similarity of pen gestures." Proceedings of the SIGCHI conference on Human Factors in Computing Systems. ACM, 2000.
Publication Link: http://dl.acm.org/citation.cfm?id=332458
Notes
These authors have a lot of work in the area of gesture recognition using different form factors and inputs. This paper discusses some of the background for their work, examining visual similarity of gestures and how this may be determined.
Friday, September 18, 2015
Extra Credit - Book Chapter Feedback
The latest reading is a draft of a document written by Dr. Hammond, and she would like to receive as much feedback on it as possible. As you're reading it this weekend, note your comments, ideas, and suggestions and then when you're finished, add them to the Google Doc shared below.
There will be a small amount of extra credit associated with each unique suggestion or correction you provide. That is, if you've found a typo, but someone else already has commented on it, you would not receive credit for that correction but any new one would be worth credit.
In addition to finding typos, if there is a part of the text that is confusing or you think could be explained in a better way, suggest it. If there is an image you think should be changed, you could mention that too. You can also contribute your own images if you have an idea for a better one or a new one that will illustrate a concept well.
You must be logged into your TAMU account to edit this document. Add your feedback below the others ahead of you. Try to only add when you have everything you want to put in the document; it's better not to be making a bunch of edits above the others.
This is only extra credit. It is not required. It will be at least a couple of weeks before this is due, but note that credit is provided first-come-first-serve. As time progresses, it will be harder to find things to comment on.
Chapter Suggestions
There will be a small amount of extra credit associated with each unique suggestion or correction you provide. That is, if you've found a typo, but someone else already has commented on it, you would not receive credit for that correction but any new one would be worth credit.
In addition to finding typos, if there is a part of the text that is confusing or you think could be explained in a better way, suggest it. If there is an image you think should be changed, you could mention that too. You can also contribute your own images if you have an idea for a better one or a new one that will illustrate a concept well.
You must be logged into your TAMU account to edit this document. Add your feedback below the others ahead of you. Try to only add when you have everything you want to put in the document; it's better not to be making a bunch of edits above the others.
This is only extra credit. It is not required. It will be at least a couple of weeks before this is due, but note that credit is provided first-come-first-serve. As time progresses, it will be harder to find things to comment on.
Chapter Suggestions
Reading 9 - Gesture Recognition Overview
Paper
Gesture Recognition Features Overview. This document is not available online. You will need to meet with me to obtain a copy.
Notes
This is an overview document of some of the common gesture recognition features, paying attention primarily to Rubine's definitions. It should be a helpful supplement to Rubine's paper (the previous reading) that will give a lot more detail about how the features work and why they were selected.
Gesture Recognition Features Overview. This document is not available online. You will need to meet with me to obtain a copy.
Notes
This is an overview document of some of the common gesture recognition features, paying attention primarily to Rubine's definitions. It should be a helpful supplement to Rubine's paper (the previous reading) that will give a lot more detail about how the features work and why they were selected.
Tuesday, September 15, 2015
Reading 8 - Rubine's Features
Paper
Rubine, Dean. Specifying gestures by example. Vol. 25. No. 4. ACM, 1991.
Publication Link: http://dl.acm.org/citation.cfm?id=122753
Notes
With this reading, we get into one of the landmark papers in sketch recognition. Although it may seem like we're just talking about template matching given the goal of the GRANDMA program, Rubine did not use other template matching algorithms. He developed several very useful sketch features that are still used today. They lay the framework for a lot of feature-based recognition, so this will be a useful paper to read. For those who are curious or wanting even more detail, you can read Rubine's entire thesis "The Automatic Recognition of Gestures" on Google Scholar, but that is by no means an assignment... just another reference should you want it.
Rubine, Dean. Specifying gestures by example. Vol. 25. No. 4. ACM, 1991.
Publication Link: http://dl.acm.org/citation.cfm?id=122753
Notes
With this reading, we get into one of the landmark papers in sketch recognition. Although it may seem like we're just talking about template matching given the goal of the GRANDMA program, Rubine did not use other template matching algorithms. He developed several very useful sketch features that are still used today. They lay the framework for a lot of feature-based recognition, so this will be a useful paper to read. For those who are curious or wanting even more detail, you can read Rubine's entire thesis "The Automatic Recognition of Gestures" on Google Scholar, but that is by no means an assignment... just another reference should you want it.
Monday, September 14, 2015
Blog Grading Policy
As promised earlier, I wanted to provide a little more detail about the blog grading. Rather than check everything at the end of the class, I think it will be better and easier for everybody to do regular checks. This will ensure that you are keeping up in the course, as well as allow me to keep a running status to ensure no one is falling too far behind.
To that effect, I plan to be looking at everyone's blogs once a week on Sunday afternoons. When I look at the blogs, I'll be checking for posts regarding the readings that were assigned up to that previous Monday. This means that Wednesday and Friday blogs are graded with the next week's Monday blog at the end of the next week.
More explicitly, this means that readings 1 through 6 are due this coming Sunday (the 20th). Readings 7 through 9 will be due the next Sunday. 10, 11, and 12 will be checked the next week, and so on. You'll have about a week to a week and a half of flexibility using this technique, but I recommend you try to keep up as much as possible.
A couple of notes:
- There are no late submissions for blogs. I will be checking for the assignments from the previous week only, and you will either receive the completion points or not. Honestly, it's not a big deal if you miss a few blogs since there are so many, which is why there's not a late policy.
- I've intentionally left the time I'll be checking vague. The reason I say Sunday afternoon is with the hope that everyone will finish everything by Saturday. If you want to try to get them in Sunday, that's fine, but it's a bit more of a gamble as to whether it will be before or after I've checked.
Email me if you have any questions or concerns regarding this policy.
Reading 7 - $1 Algorithm
Paper
Wobbrock, Jacob O., Andrew D. Wilson, and Yang Li. "Gestures without libraries, toolkits or training: a $1 recognizer for user interface prototypes." Proceedings of the 20th annual ACM symposium on User interface software and technology. ACM, 2007.
Publication Link: http://dl.acm.org/citation.cfm?id=1294238
Notes
This paper is very relevant to the template matching discussion. It's a pretty basic template matching technique and is widely used.
Wobbrock, Jacob O., Andrew D. Wilson, and Yang Li. "Gestures without libraries, toolkits or training: a $1 recognizer for user interface prototypes." Proceedings of the 20th annual ACM symposium on User interface software and technology. ACM, 2007.
Publication Link: http://dl.acm.org/citation.cfm?id=1294238
Notes
This paper is very relevant to the template matching discussion. It's a pretty basic template matching technique and is widely used.
Friday, September 11, 2015
Reading 6 - Entropy To Decide Sketch Type
Paper
Bhat, Akshay, and Tracy Hammond. "Using Entropy to Distinguish Shape Versus Text in Hand-Drawn Diagrams." IJCAI. Vol. 9. 2009.
Direct Link: http://ijcai.org/papers09/Papers/IJCAI09-234.pdf
Notes
Here we see how entropy may be used to determine shape types. It's a pretty interesting paper, and hopefully it will give some idea about the considerations made regarding hand-writing and sketch recognition.
Bhat, Akshay, and Tracy Hammond. "Using Entropy to Distinguish Shape Versus Text in Hand-Drawn Diagrams." IJCAI. Vol. 9. 2009.
Direct Link: http://ijcai.org/papers09/Papers/IJCAI09-234.pdf
Notes
Here we see how entropy may be used to determine shape types. It's a pretty interesting paper, and hopefully it will give some idea about the considerations made regarding hand-writing and sketch recognition.
Wednesday, September 9, 2015
Reading 5 - Trainable Symbol Recognizer
Paper
Kara, Levent Burak, and Thomas F. Stahovich. "An image-based, trainable symbol recognizer for hand-drawn sketches." Computers & Graphics 29.4 (2005): 501-517.
Publication Link: http://www.sciencedirect.com/science/article/pii/S0097849305000853
Notes
This paper should be useful to give a little more algorithmic background.
Kara, Levent Burak, and Thomas F. Stahovich. "An image-based, trainable symbol recognizer for hand-drawn sketches." Computers & Graphics 29.4 (2005): 501-517.
Publication Link: http://www.sciencedirect.com/science/article/pii/S0097849305000853
Notes
This paper should be useful to give a little more algorithmic background.
Monday, September 7, 2015
Course Notes Up to Covariance
We've covered a lot of information in the last few classes, and with no definitive reference, I know some students are finding it hard to keep up. Especially for those of you who were not present due to joining the course late or traveling, I imagine it's hard to figure out the material. For this reason, I've put together a few pages summarizing what we've covered so far. I will probably not do this for very many lectures, but if the material is particularly challenging, further summaries may appear in the future.
Course Notes - Classification to Covariance
Course Notes - Classification to Covariance
Worksheet 1 Link
For those who don't have it or would like to print a fresh copy for practice, here is the link to Worksheet 1 which we went through in class:
Worksheet 1
Worksheet 1
Reading 4 - K Sketch
Paper
Davis, Richard C., Brien Colwell, and James A. Landay. "K-sketch: a'kinetic'sketch pad for novice animators." Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. ACM, 2008.
Publication Link: http://dl.acm.org/citation.cfm?id=1357122
Notes
This is another fairly simple paper. It gives you a look at another sketching domain, which will hopefully start giving you ideas for your own project. Again, just give it a short overview reading and put up a few sentences on your blog.
Davis, Richard C., Brien Colwell, and James A. Landay. "K-sketch: a'kinetic'sketch pad for novice animators." Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. ACM, 2008.
Publication Link: http://dl.acm.org/citation.cfm?id=1357122
Notes
This is another fairly simple paper. It gives you a look at another sketching domain, which will hopefully start giving you ideas for your own project. Again, just give it a short overview reading and put up a few sentences on your blog.
Friday, September 4, 2015
Reading 3 - iCanDraw?
Paper
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.
Publication Link: http://dl.acm.org/citation.cfm?id=1753459
Notes
This is a fun paper about a software program that grew out of a sketch recognition course project. It should be helpful to you since it will introduce some other recognition algorithms as well as show another problem domain for sketch recognition.
Presenter Notes
This paper includes some information about the comparison algorithm and the error evaluation equation. It will be most beneficial to the other students to focus on this content in the presentation to explain the algorithm more clearly.
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.
Publication Link: http://dl.acm.org/citation.cfm?id=1753459
Notes
This is a fun paper about a software program that grew out of a sketch recognition course project. It should be helpful to you since it will introduce some other recognition algorithms as well as show another problem domain for sketch recognition.
Presenter Notes
This paper includes some information about the comparison algorithm and the error evaluation equation. It will be most beneficial to the other students to focus on this content in the presentation to explain the algorithm more clearly.
Wednesday, September 2, 2015
Presenter Schedule
Presentations will begin to be planned further in advance, so it is useful to have a reference to the speakers. The link below leads to a spreadsheet showing (a) the date of the presentations, (b) the reading assignment, and (c) the individual assigned to present. Please reference it in the future:
Presenter Schedule
Presenter Schedule
Reading 2 - Mechanix Algorithms
Paper
Field, Martin, et al. "Sketch recognition algorithms for comparing complex and unpredictable shapes." IJCAI. 2011.
Direct Link: http://www.researchgate.net/profile/Tracy_Hammond/publication/220812144_Sketch_Recognition_Algorithms_for_Comparing_Complex_and_Unpredictable_Shapes/links/0deec529f76e510edf000000.pdf
Notes
This paper delves into more detail about the recognition algorithms than the previous one. It is a helpful follow-up since it should give you a deeper understanding of how Mechanix works.
Presenter Notes
While this paper provides a good description of both the Hausdorff method and the truss recognizer, you will want to focus on the Hausdorff recognizer in your talk. Since there are two papers on Mechanix, this has to be slightly more coordinated than most presentations will be (hence the presence of these special notes). The AI Magazine paper will be used to provide most of the information for the truss recognizer talk, and this paper will be used to give the Hausdorff method presentation. Contact me if you have any questions.
Field, Martin, et al. "Sketch recognition algorithms for comparing complex and unpredictable shapes." IJCAI. 2011.
Direct Link: http://www.researchgate.net/profile/Tracy_Hammond/publication/220812144_Sketch_Recognition_Algorithms_for_Comparing_Complex_and_Unpredictable_Shapes/links/0deec529f76e510edf000000.pdf
Notes
This paper delves into more detail about the recognition algorithms than the previous one. It is a helpful follow-up since it should give you a deeper understanding of how Mechanix works.
Presenter Notes
While this paper provides a good description of both the Hausdorff method and the truss recognizer, you will want to focus on the Hausdorff recognizer in your talk. Since there are two papers on Mechanix, this has to be slightly more coordinated than most presentations will be (hence the presence of these special notes). The AI Magazine paper will be used to provide most of the information for the truss recognizer talk, and this paper will be used to give the Hausdorff method presentation. Contact me if you have any questions.
Reading 1 - Mechanix Overview
Paper
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
Notes
This is a good overview paper about Mechanix. It's a useful first reading since it covers a lot of information about the program, including some modes we have not discussed in the class.
Presenter Notes
For the two Mechanix papers, Dr. Hammond would like the presenters to discuss primarily the two algorithms outlined in this paper and the next. In this paper, the most-discussed algorithm is the truss recognition method using shared edges and searching. Thus, for the presenter of this paper, while you can briefly address the other material about Mechanix with an overview of its capabilities, focus on the truss recognition algorithm. Contact me if you have any questions.
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
Notes
This is a good overview paper about Mechanix. It's a useful first reading since it covers a lot of information about the program, including some modes we have not discussed in the class.
Presenter Notes
For the two Mechanix papers, Dr. Hammond would like the presenters to discuss primarily the two algorithms outlined in this paper and the next. In this paper, the most-discussed algorithm is the truss recognition method using shared edges and searching. Thus, for the presenter of this paper, while you can briefly address the other material about Mechanix with an overview of its capabilities, focus on the truss recognition algorithm. Contact me if you have any questions.
General Guidelines Regarding Papers
In this course, papers have two graded components.
Blog Posts
One blog post should be made for every paper assigned in the course. Because this is done with every paper, it is not meant to be very burdensome. You are asked to take a small amount of time to skim over the paper quickly, just enough to get the key points so that you know what the paper is about. From this information, make a blog post according to the following format:
Presentations
Everybody will need to present on at least one paper in this course. Presentations are given at the beginning of class from the front of the room. You should have some slides to augment your talk.
Presentations will be no more than 7 minutes in length, saving 3 minutes at the end for questions. This allows for a maximum of 10 minutes per speaker.
This is primarily a completion grade since this is not a public speaking course. However, it will be useful practice for you since you will undoubtedly (and may have already) present many research-related talks. So please prepare appropriately so that you feel you can comfortably deliver a description of the paper and answer reasonable questions.
Blog Posts
One blog post should be made for every paper assigned in the course. Because this is done with every paper, it is not meant to be very burdensome. You are asked to take a small amount of time to skim over the paper quickly, just enough to get the key points so that you know what the paper is about. From this information, make a blog post according to the following format:
Citation
Include author list, paper name, publication information, and link to the paper (either a PDF or publication page). This should be some standard, readable citation. Google scholar can generate these for you if you search the paper name.
Summary
Your high-level description of the paper. This does not need to be very long. Just write a few sentences or a paragraph explaining enough about the paper that you will be able to remember it for later reference. This paragraph should help you decide if you want to include the paper in your final report.
Discussion
Your impression of the paper. This can include your thoughts on the method, what ideas it gave you, or anything else related to the paper. This should be useful for you to peruse later as it may help give you ideas for your own project or remind you of some other, similar work you wanted to explore.
These posts will be mostly graded on completion, but you must put in sufficient effort to make it apparent that you have quickly read through the paper and would be able to talk intelligently about it.
Presentations
Everybody will need to present on at least one paper in this course. Presentations are given at the beginning of class from the front of the room. You should have some slides to augment your talk.
Presentations will be no more than 7 minutes in length, saving 3 minutes at the end for questions. This allows for a maximum of 10 minutes per speaker.
This is primarily a completion grade since this is not a public speaking course. However, it will be useful practice for you since you will undoubtedly (and may have already) present many research-related talks. So please prepare appropriately so that you feel you can comfortably deliver a description of the paper and answer reasonable questions.
Subscribe to:
Posts (Atom)