Framework for research on detection classifiers

Object Detection, WaldBoost, AdaBoost, Viola&Jones, LRD, LRP

Detecting objects belonging to a certain visual class in images is one of the fundamental tasks of computer vision. Some of the target applications in this field are for example detection of faces for surveillance systems and photo gallery software, detection of traffic signs for driver assistant systems and also licence plate localization for traffic monitoring systems. An approach which solves a large subset of the possible detection tasks is scanning images for specific 2D patterns using classifiers. Our group focuses on improving these detection classifiers and on adapting them to other computing platforms beside CPU. The implementations of classification engines on GPU, GP-GPU and FPGA which we have created provide high precision of detection and are able to process high-resolution video signals in real-time. During the development of new classifiers, we have created a framework which provides good support for effective development and testing of novel machine learning algorithms, image features and data sets. We provide this framework for research and educational purposes without any fee. Fee is required when using the framework for commercial purposes. We provide a  [[a href="http://pcl210-01.fit.vutbr.cz/trenovadlo/index.php"]] simplified web interface [[/a]] which allows you to train detection classifiers using your own datasets and which also allows you to evaluate the performance of the created classifiers on your own testing datasets.

http://pcl210-01.fit.vutbr.cz/trenovadlo/index.php