Fotowoosh 3-D images

Automatic Photo Pop-up 3D images

Al-Yaman Awad MAT267 FOOTNOTE 18

Supervised by Professor Naala Brewer

As the world becomes more technologically advanced, the required pace of learning is increasing to keep up with the vast amounts of knowledge. The learning curve is increasing and with it students need more and more tools to help them attain an advanced comprehensive perspective on many early but fundamental concepts. One of the most fundamental mathematical concepts that numerous students experience problems with is 3-D coordinal systems and graphs. This concept is taught extensively in Calculus 3 after a 2-D approach in Calculus 2. While completing work and passing exams may be feasible for students, an advanced level of comprehension of 3-D systems is somewhat difficult to ascertain, and may feel overwhelming at first. Being able to mentally draw what the graph for a hyperbolic paraboloid or hyperboloid of two sheets is extremely difficult even for engineers unless they have had a solid understanding of how to approach these types of functions. Most of the problems can be eliminated by addressing not only the method by which people perceive equations but the world in general.

After speaking with some older friends early on during the semester about the differences between calculus 2 and 3, we came to the conclusion that just as long as you can describe and imagine the graphs you are drawing you should be fine. One of them led me to a new program called “fotowoosh”, an automatic pop-up 3D image. This new program uses mathematic principles to transforms 2D images that you would take with a digital camera into 3D by morphing the angle of perspective and stretching it across a variable axis. That axis is the vertical axis that gives the images we see using our eyes depth (How does it work?). Although the math behind this software deals with linear algebra and algorithms, the application greatly exceeds these two concepts. By seeing these images broken down and the perspective changed, Calculus students can see first hand what 3D graphs actually are. It enables students to have a new outlook on the way they see the world around them, allowing them to more extensively comprehend what the addition of a z axis does.

While there are a few advanced programs already in use by universities across the world like MATLAB and Maple that construct 3D images using an input function, the approach of Fotowoosh is what is particularly appealing in addition to its user friendly design. Rather than just drawing graphs and shapes that students can find in their textbooks, it brings the concept of 3D imaging into to everyday life. Dr. Hoiem created the program with the help of Prof. Efros and Prof. Hebert at Carnegie Mellon University in Philadelphia as an ongoing search for ways to transform 2D images into 3D. Carnegie Melon is currently leading the country in 3D computer design technology and research. The university ranks on the top 25 best universities in all of America (America's Best Colleges 2008).

What the software does is revolutionary since it is the first software to automatically create the pop-up 3D image. While manual methods do produce better results, it simply requires too much extensively complicated work, which the average person will be unable to complete. The goal was to produce simple, piecewise planar models, using images of outdoor scenes. What the program does in a gist is label in the image where the ground, vertical, and sky of the image are. It puts a simple label on each of these that will enable the program to determine where to cut and fold the image. After the system has done this, it folds and stretches the image and textures it. By doing this a pop-up 3D image is created similar to a child’s pop-up book.

Fotowoosh recently released their first public beta through Facebook as an application. It allows the user to upload images they currently have saved in their profile. The beta automatically “wooshes” your image and creates a video for the pop-up 3-D image. While the current beta features a success rate of about 35%, the program will only get better and better.

Demo video:

http://www.cs.uiuc.edu/homes/dhoiem/projects/popup/popup_movie_450_250.mp4

Derek Hoiem’s webpage: (contains tons of interesting projects and information)

http://www.cs.cmu.edu/~dhoiem/

Derek Hoiem’s Automatic Photo Pop-Up homepage: (download presentation file for great videos)

Steps in Creating a 3-D Image
Input
Labeled

Ready for cut & fold

Green color labels the ground. Red labels the vertical space. Purple labels the sky.

3D image

Work cited

Thanks:

This Honors Footnote 18 presentation was made possible with the help of Derek Hoiem’s research and work, and his personal help. Also, I would like to thank Prof. Efros and Prof. Hebert for their work in the project. Lastly, I would like to thank Carnegie Mellon University for funding research like this among other important work.