Ensembles of strong learners for multi-cue classification

Abstract

Real world heterogeneous scenes contain objects of a large variety of forms, surfaces, colors and textures, thus multi-modal approaches are needed to deal with their challenges. A promising method of combining various sources of information are ensemble methods which allow on the fly integration of classification modules, specific to a single sensor modality, into a classification process. These modular and extensible approaches have the advantage that they do not require that a single method copes with every eventuality, but combine existing specialized methods to overcome their weaknesses.In addition, the rapid growth of the perception field means that comparing, evaluating, sharing and combining the available approaches becomes increasingly relevant. In this article we describe a novel training strategy for ensembles of strong learners that not only outperform the best member but also the best classifier trained on the concatenation of features.The method was evaluated using a large RGBD dataset containing Kinect scans of 300 objects and special use-cases are presented that highlight how ensemble learning can be used to improve classification results.

Improving object classification robustness in RGB-D using adaptive SVMs

“…We considered adaptation with SVMs only, as they are very widely used, but other classifiers might perform better. However, based on [21], SVMs with RBF kernels are already quite good when compared to other standard classifiers.…”

“…Although in our prospective work we also plan to investigate other classification algorithms for the problem addressed in this article, our decision is motivated by the conclusions from the state-of-the-art, e.g., [21]. In this article we mainly study the impact of different feature extraction techniques on our adaptive methodology.…”

Grounded object individuation by a humanoid robot

“…In contrast, a wide variety of methods have been developed that allow robots to recognize previously observed objects. The majority of these methods use 2D and 3D visual features (see [14], [15], [7], [9], [11]). Other vision-based approaches have also been proposed for finding image regions from multiple views that contain the same object [16].…”

“…In contrast, most methods used by robots to recognize objects start with a fixed object representation in which the robot's training data is labeled with one of a finite number of object identities (see [4], [5], [6], [7], [8], [9], [10], [11] for a representative sample of such approaches). These methods implicitly make the assumption that the object individuation task has already been solved.…”

Multimodal blending for high-accuracy instance recognition

“…One simple method is to learn an affine combination of the scores obtained through each modality. Another method that extends this idea is feature weighted linear stacking [15], [16]. We evaluate both methods for the purpose of generating hypothesis detections.…”