SciTech

Researchers hope to solve ‘fat-finger’ problem with app

By Nivedita Chopra  |  Aug 18, 2013
In DrawAFriend, one player attempts to draw the outline of a celebrity or a mutual friend with the other participant while the other player attempts to guess the identity of the person. Above are two screenshots from the app. (credit: Courtesy of Alex Limpaecher) In DrawAFriend, one player attempts to draw the outline of a celebrity or a mutual friend with the other participant while the other player attempts to guess the identity of the person. Above are two screenshots from the app. (credit: Courtesy of Alex Limpaecher) In DrawAFriend, one player attempts to draw the outline of a celebrity or a mutual friend with the other participant while the other player attempts to guess the identity of the person. Above are two screenshots from the app. (credit: Courtesy of Alex Limpaecher) In DrawAFriend, one player attempts to draw the outline of a celebrity or a mutual friend with the other participant while the other player attempts to guess the identity of the person. Above are two screenshots from the app. (credit: Courtesy of Alex Limpaecher)

If you’ve ever tried drawing or writing on a touchscreen, you may have noticed that the lines drawn on the screen aren’t exactly where you wanted them to be. This “fat finger” problem arises because your finger is pretty large or “fat” compared to the small touchscreen. Recently, a team of computer scientists from Carnegie Mellon and Microsoft Research developed an app that automatically corrects drawing strokes while preserving the user’s artistic style.

This app, DrawAFriend, is currently available to download from the Mac App Store. It is a two-player game in which one player draws a picture of a celebrity or mutual Facebook friend while the other guesses the name of the person drawn in a Hangman-like fashion. The app automatically corrects users’ strokes, often without them noticing. “Our goal was to make it invisible to the user, so people wouldn’t even be aware the correction is taking place,” said Alex Limpaecher, a Ph.D. student in computer science, in a university press release.

The team consisted of Limpaecher, computer science professor Adrien Treuille, Ph.D. student in computer science Nicolas Feltman, computer science professor Adrien Treuille, and Michael Cohen, a principal researcher in Microsoft Research’s Interactive Visual Media Group.

Initially, the main goal of the app was to gather data, which it did by encouraging users to draw sketches of Angelina Jolie, Brad Pitt, and other celebrities for their friends or other users to guess. In its first week, the game generated 1,500 images a day, creating a huge database of drawings with stroke-by-stoke information. The game is still operational, and the database now exceeds 17,000 entries. This huge database of drawings allowed the team to analyze the drawings and come up with a stroke-correction algorithm. They applied the algorithm to future drawings on the app, allowing current users to create more precise sketches.

This is one of many problems that have been solved using big data. “Big data” refers to the collection of a large amount of specific data — in this case, drawings — to observe patterns in the data and solve a problem. The larger the data set, the more precise the patterns become and the better the solutions will be. In this case, a large database of drawings was essential for the researchers to correlate the strokes in the images, thus creating the algorithm for eliminating the noise associated with the fat finger problem.

In a press release, Treuille said that the drawing assistance app is just one example of how big data can be used to enhance drawing and writing on touchscreens and even provide insight into art and perception. The trick has been to create drawing databases large enough to leverage, which the research team did by developing DrawAFriend.

According to Limpaecher, the idea for the game stemmed from earlier projects and a need to address the difficulties of gathering data involving drawings. Limpaecher had personally dealt with this problem. In a previous project, his team collected the data set from an artist. Not only did it take a long time to get the drawings; they were able to obtain only 154 images, not nearly enough to support a drawing assistance app. The goal here was “to create a drawing helper that used other people’s drawings to help you draw,” Limpaecher said via email.

At the moment, the app only helps people who are already adequate at drawing on a touchscreen but may need some assistance. Limpaecher says that in the future, he would like to add features that can help people at any talent level.