3D texture classification under varying viewpoint and illumination has been a vivid research topic, and many methods have been developed. We performed a series of experiments which show that standard approaches to evaluation of these methods lead to overestimation of classifier performance and distort experimental findings. In this paper we propose a methodology for more reliable evaluation of the texture classification algorithms. We show that such methodology gives a more realistic assessment of classifier performance.
COBISS.SI-ID: 7683924
In this paper we present a framework for visual context aware object detection. The concept is based on a sparse coding of contextual features, which are based on geometry and texture. In addition, bottom-up saliency and object co-occurrences are exploited, to define auxiliary visual context. To integrate the individual contextual cues with a local appearance-based object detector, a fully probabilistic framework is established, which is based on modelling the underlying conditional probabilities between the different cues via kernel density estimation.
COBISS.SI-ID: 7684180
The book Cognitive Systems addresses the function and the development of artificial cognitive systems from different perspectives. The chapter on Multi-modal learning addresses the problem of learning on different levels; from the learning in interaction with a human tutor, to the learning in interaction with the environment and objects there-in. We pay a special attention to representations that are used for modelling visual concepts, and to algorithms used for building these representations, which are mainly incremental, utilising different levels of the system's autonomy.
COBISS.SI-ID: 8305236
In this work we present a new approach to interest point detection. Different types of features in images are detected by using a common computational concept. The proposed approach consider the total variability of local regions. Results obtained on a wide variety of image sets compare favourably with the results obtained by the leading interest point detectors from the literature. The proposed approach gives a rich set of highly distinctive local regions that can be used for object recognition and image matching.
COBISS.SI-ID: 41304418
Visualizing music in a meaningful and intuitive way is a challenge. Our aim is to visualize music by interconnecting similar aspects in music and in visual perception. As concurrent tones are not perceived as entirely separate, but also as a whole, we present a novel method for visualizing a group of concurrent tones with one colour for the whole group. The basis for calculation of colour is the assignment of key spanning circle of thirds to the colour wheel.
COBISS.SI-ID: 8061268