We have proposed a novel approach to online estimation of generative models, which is based on probability density estimation. As the theoretical framework we have chosen the kernel density estimation (KDE). The method maintains and updates a non-parametric model of the observed data, from which the KDE can be calculated. We propose an online bandwidth estimation approach and a compression/revitalization scheme which maintains the KDE's complexity low. We compare the proposed online KDE to the state-of-the-art approaches on examples of estimating stationary and non-stationary distributions, and on examples of classification. The results show that the online KDE outperforms or achieves a comparable performance to the state-of-the-art and produces models with a significantly lower complexity while allowing online adaptation.
COBISS.SI-ID: 8289876
We have proposed a new approach to tracking using a hierarchical visual model. We developed a new probabilistic model for tracking nonrigid objects. The model is composed of two layers. The top layer consists of a global information about the appearance of the object, while the lower layer contains the object’s local information. The local layer consists of a vocabulary of basic visual elements. These visual elements are weakly coupled in a constellation that can deform during the tracking. For this purpose, we have developed a new statistical model and have proposed an optimization method, which efficiently adapts to the target’s appearance through significant deformation and partial occlusion. On the one hand, the proposed model is a test-bed for application of parts of theory of the hierarchical models to the case of object tracking. On the other hand, we developed a robust tracking algorithm that will allow us to capture large datasets of articulated objects. These will be used as an input to the object category detection hierarchy that we are developing.
COBISS.SI-ID: 8801364
Text detection and optical character recognition (OCR) in images of natural scenes is a fairly new computer vision area but yet very useful in numerous applicative areas. Although many implementations gain promising results, they are evaluated mostly on the private image collections that are very hard or even impossible to get. Therefore, it is very difficult to compare them objectively. Since our aim is to help the research community in standardizing the evaluation of the text detection and OCR methods, we present CVL OCR DB, a public database of annotated images of text in diverse natural scenes, captured at varying weather and lighting conditions. All the images in the database are annotated with the text region and single character location information, making CVL OCR DB suitable for testing and evaluating both text detection and OCR methods. Moreover, all the single characters are also cropped from the original images and stored individually, turning our database into a huge collection of characters suitable for training and testing OCR classifiers.
COBISS.SI-ID: 8697428