BieColl - Bielefeld eCollections
Not a member yet
1004 research outputs found
Sort by
GalSOM - Colour-Based Image Browsing and Retrieval with Tree-Structured Self-Organising Maps
Get PDFThis paper describes an image browsing and retrieval application called GalSOM. Bitmap images are described by their colour histograms and sorted using an improved variant of the tree-structured self-organising map (TS-SOM) algorithm. The advantages of using such a system are discussed in detail, and their application to the problem of image theft detection is proposed
The self-organizing map as a visual neighbor retrieval method
Get PDFWe have recently introduced rigorous goodness criteria for information visualization by posing it as a visual neighbor retrieval problem, where the task is to find proximate high-dimensional data based only on a low-dimensional display. Standard information retrieval criteria such as precision and recall can then be used for information visualization. We introduced an algorithm, Neighbor Retrieval Visualizer (NeRV), to optimize the total cost of retrieval errors. NeRV was shown to outperform alternative methods, but the SOM was not included in the comparison. In empirical experiments of this paper the SOM turns out to be comparable to the best methods in terms of (smoothed) precision but not on recall. On a related measure called trustworthiness, the SOM outperforms all others. Finally, we suggest that for information visualization tasks the free parameters of the SOM could be optimized for information visualization with cross-validation
Variable-Density Self-Organizing Map for Incremental Learning
No full textWe propose a new incremental learning method of Self-Organizing Map. Basically, there are three problems in the incremental learning of Self-Organizing Map: 1. depletion of neurons, 2. oblivion of training data previously given, 3. destruction of topological relationship among training samples. Weight-fixed neurons and weight-quasi-fixed neurons are very effective for the second problem. However the other problems still remain. Therefore, we improve the incremental learning method with weight-fixed neurons and weight-quasi-fixed neurons. We solve the problems by introducing a mechanism to increase the number of neurons effectively in the incremental learning process
Video Summarization with SOMs
Get PDFVideo summarization is a process where a long video file is converted to a considerably shorter form. The video summary can then be used to facilitate efficient searching and browsing of video files in large video collections. The aim of successful automatic summarization is to preserve as much as possible from the essential content of each video. What is essential is of course subjective and also dependent on the use of the videos and the overall content of the collection. In this paper we present an overview of the SOM-based methodology we have used for video summarization, which analyzes the temporal trajectories of the best-matching units of frame-wise feature vectors. It has been developed as a part of PicSOM, our content-based multimedia information retrieval and analysis framework. The video material we have used in our experiments comes from NIST's annual TRECVID evaluation for content-based video retrieval systems
SVM-based Transfer of Visual Knowledge Across Robotic Platforms
Get PDFThis paper presents an SVM--based algorithm for the transfer of knowledge across robot platforms aiming to perform the same task. Our method exploits efficiently the transferred knowledge while updating incrementally the internal representation as new information is available. The algorithm is adaptive and tends to privilege new data when building the SV solution. This prevents the old knowledge to nest into the model and eventually become a possible source of misleading information. We tested our approach in the domain of vision-based place recognition. Extensive experiments show that using transferred knowledge clearly pays off in terms of performance and stability of the solution
Component Selection for the Metro Visualisation of the Self-Organising Map
Get PDFSelf-Organising Maps have been used for a wide range of clustering applications. They are well-suited for various visualisation techniques to offer better insight into the clustered data sets. A particularly feasible visualisation is the plotting of single components of a data set and their distribution across the SOM. One central problem of the visualisation of Component Planes is that a single plot is needed for each component; this understandably leads to problems with higher-dimensional data. We therefore build on the Metro Visualisation for Self-Organising Maps which integrates the idea of Component Planes into one illustration. Higher-dimensional data sets still pose problems in terms of overloaded visualisations - component selection and aggregation techniques are highly desirable. We therefore propose and compare two methods, one for the aggregation of correlated components, one for the selection of the components most feasible for visualisation for a given clustering
Topographic Processing of Relational Data
Get PDFRecently, batch optimization schemes of the self-organizing map and neural gas have been modified to allow arbitrary distance measures.This principle is particularly suitable for complex applications where data are compared by means of problem-specific, possibly discrete metrics such as protein sequences. However, median variants do not allow a continuous update of prototype locations and their capacity is thus restricted. In this contribution, we consider the relational dual of batch optimization which can be formulated in terms of pairwise distances only such that an application to arbitrary distance matrices becomes possible. For SOM, a direct visualization of data is given by means of the underlying (euclidean or hyperbolic) lattice structure. For NG, pairwise distances of prototypes can be computed based on a given data matrix only, such that subsequent mapping by means of multidimensional scaling can be applied
Active Vision-based Localization For Robots In A Home-Tour Scenario
Get PDFSelf-Localization is a crucial task for mobile robots. It is not only a requirement for auto navigation but also provides contextual information to support human robot interaction (HRI). In this paper we present an active vision-based localization method for integration in a complex robot system to work in human interaction scenarios (e.g. _home-tour_) in a real world apartment. The holistic features used are robust to illumination and structural changes in the scene. The system uses only a single pan-tilt camera shared between different vision applications running in parallel to reduce the number of sensors. Additional information from other modalities (like laser scanners) can be used, profiting of an integration into an existing system. The camera view can be actively adapted and the evaluation showed that different rooms can be discerned
Binarized Eigenphases for Limited Memory Face Recognition Applications
Get PDFMost of the algorithms proposed for face recognition involve considerable amount of calculations, and hence they can not be used on devices of limited memory constraints. In this paper, we propose a novel solution for efficient face recognition problem for the systems that utilize low memory devices. The new technique applies the principal component analysis to the binarized phase spectrum of the Fourier transform of the covariance matrix constructed from the MPEG-7 Fourier Feature Descriptor vectors of the images. The binarization step that is applied to the phases adds many interesting advantages to the system. It will be shown that the proposed technique maximizes the recognition rate while achieving substantial savings in computational time, when compared to other known systems
Computational Attention for Event Detection
Get PDFThis article deals with a biologically-motivated three-level computational attention model architecture based on the rarity and the information theory framework. It mainly focuses on low-level and medium-level steps and their application in pre-attentive detection of tumours in CT scans and unusual events in audio recordings