Object classification by fusing SVMs and Gaussian mixtures

Abstract

We present a new technique that employs support vector machines (SVMs) and Gaussian mixture densities (GMDs) to create a generative/discriminative object classification technique using local image features. In the past, several approaches to fuse the advantages of generative and discriminative approaches were presented, often leading to improved robustness and recognition accuracy. Support vector machines are a well known discriminative classification framework but, similar to other discriminative approaches, suffer from a lack of robustness with respect to noise and overfitting. Gaussian mixtures, on the contrary, are a widely used generative technique. We present a method to directly fuse both approaches, effectively allowing to fully exploit the advantages of both. The fusion of SVMs and GMDs is done by representing SVMs in the framework of GMDs without changing the training and without changing the decision boundary. The new classifier is evaluated on the PASCAL VOC 2006 data. Additionally, we perform experiments on the USPS dataset and on four tasks from the UCI machine learning repository to obtain additional insights into the properties of the proposed approach. It is shown that for the relatively rare cases where SVMs have problems, the combined method outperforms both individual ones.

Classification of urban areas from GeoEye-1 imagery through texture features based on Histograms of Equivalent Patterns

“…In fact, combining information from different feature vectors or classifiers represents an important research line in the field of image classification [Kittler et al, 1998;Segl et al, 2003;Permuter et al, 2006;Deselaers et al, 2010]. In theory, the integration of different and independent sources of information should improve the classification accuracy [Kittler et al, 1998].…”

3D model-based still image object categorization

“…Some approaches based on supervised machine learning methods are proposed in [13], [14], [15]. Even if the supervised approaches may be highly accurate [16], they can often suffer from overfitting [17]. Another limitation comes from the need of sufficiently large training sets with already classified objects.…”

“…The goal is to automatically identify the semantics of the image/video objects present in the content, with the help of computer vision methods. To achieve this goal, the majority of state of the art methods [1], [2] exploit some machine learning techniques. Such approaches need to perform a learning stage on sufficiently large databases.…”

3D Model-Based Sematic Labeling of 2D Objects

“…Once the function is determined, it can be applied for each unknown object in order to determine to which category the object belongs. The supervised ML techniques may be highly accurate [3]. However, it may happen that the function is too appropriate for the training set and inadequate for new objects [4].…”