Improving Multi-face Detection Using Bimodality Approach on Viola Jones

Multi-face detection has a different level of difficulty with single face detection because it requires the ability to detect at various scales/sizes. Viola Jones can detect front-facing multi faces very well with AdaBoost's support, one of the most popular algorithms for creating a strong classifier in detection, even though there are still problems with the rate of false positives appearing. This study will discuss the optimization of Viola Jones performance by modifying the post-stage of Viola Jones classifier using the concept of a Bimodality approach that works in a split and merge. The dataset used is in the form of an image file that will be tested using an application built using the ACCORD Library which has been modified according to the proposed method. Tests will be carried out in stages on several color modes and also with and without applying the proposed method to then compare the test results. The test results show an increase in accuracy from 68% to 75% compared to the original method.