A Novel Search for Face Attractiveness Using Reverse Correlation Methods and Web-Based Visual Experiments

Overview

Objectives/Goals With our research, we can obtain a more complete picture of the brain and its involvement in perception. Neuroscientists can use our model to understand people#s perception of beauty, and plastic surgeons and developers in the beauty industry can use this technique to make people look more attractive. Methods/Materials This Web-based experiment is set up on a Linux computer using the Apache Web server and a mySQL database and is written in Perl. Observers created an account,logged in, and selected an experiment. They rated the attractiveness (from 1 to 10, with 10 being the highest) of 100 stimuli, either male or female images, randomly generated from the data file on-the-fly. Finally, to analyze the test data, we imported the user information and responses from the database to MATLAB and EXCEL for further mathematical and statistical analysis. Results By determining a correlation between the observer#s response and a filtered signal pattern, we created a kernel. After adding the kernel to the base image using reverse correlation, we produced a more attractive face than the original one. Out of the 4,092 pixels representing a human face, only 134 key pixels contribute to an attractive face. Conclusions/Discussion The reverse correlation technique and classification image (kernel) are effective methods for studying face attractiveness. Reverse correlation directly estimates the areas of an image that observers used to reach their decisions. Since this method is based on one model, it can determine the significant components that yield attractiveness. This Web-based experiment offered significant advantages to traditional lab experiments. Outside of the U.S., we gathered data from participants all over the world in countries like the United Kingdom, Greece, Spain, Australia, and the United Arab Emirates. Online experiments allow for hundreds of users to participate in the experiment simultaneously at minimal costs. As high bandwidth becomes more readily available, we can incorporate more features such as dynamic filtered signals, movie scripts, and colorful images into online experiments. Our goal is to investigate spatial sampling strategies used by observers in choosing between an unattractive and attractive face. Finally, we can explore possibilities of the model in other fields of studies in Neuroscience and the practical applications of this technique in the beauty industries.

Competition history

• CSEF 2002 Mathematics & Software · Entry S1210

Resources

• CSEF project summary ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

• All Computational Sciences projects
• CSEF Computational Sciences projects
• Every CSEF 2002 project