A Compound Technique for Multiple Objects Detection Based on Markov Clustering Networks and Viola-Jones Algorithm
Teddy Hendra, Riccardo Spolaor, Zhenxiang Chen · 2019
In this paper, we present a computer vision technique for detecting multiple objects simultaneously. In particular, our proposal specifically aims to detect the presence of faces and text within an image. To achieve this goal, our technique relies on Markov Clustering Networks (MCN) and Viola-Jones algorithm. These algorithms perform object detection without prior consideration of the external factors, such as the image size, gradient, orientation, and characters. While we identify the presence of such objects within a single image, our technique is also able to detect multiple lookalike faces (e.g. doll, mask) and multiple textual objects with different orientation. As a result of the experimental evaluation, our proposed technique able to detect image with frontal face and text with 90% accuracy and around 80% accuracy for images with non-frontal face and text. Finally, our technique achieves a new state of the art performance with an average precision of 88%, performs real-time inference with a frame rate of above 50 Frame Per Second (FPS) and has a faster computation time when compared with other techniques.