This paper presents a new method for performance evaluation of local detectors on non-planar scenes. The framework extends the detector evaluation of Mikolajczyk and Schmid to non-planar scenes. The basic idea for est...
Modelling and reconstruction of tubular objects is a known problem in computergraphics. For computer aided surgical planning the constructed geometrical models need to be consistent and compact at the same time, whic...
详细信息
ISBN:
(纸本)8090310079
Modelling and reconstruction of tubular objects is a known problem in computergraphics. For computer aided surgical planning the constructed geometrical models need to be consistent and compact at the same time, which known approaches cannot guarantee. In this paper we present a new method for generating compact, topologically consistent, 2-manifold surfaces of branching tubular objects using a two-stage approach. The proposed method is based on connection of polygonal cross-sections along the medial axis and subsequent re nement. Higher order furcations can be handled correctly. Copyright UNION Agency Science Press.
In this paper we introduce a framework for automated text recognition from images. We first describe a simple but efficient text detection and recognition method based on analysis of Maximally Stable Extremal Regions ...
详细信息
Ray tracing on massively parallel hardware allows for the computation of images with a high visual quality in an increasingly short time. However, apping the computations to such architectures in an efficient manner i...
详细信息
The creation of clutter-free orthophotos and 3D GIS data bases is desireable. Cars parked in the streets or driving on the road are of little interest in an orthophoto or in a GIS. We are demonstrating the ability of ...
详细信息
ISBN:
(纸本)9781605604046
The creation of clutter-free orthophotos and 3D GIS data bases is desireable. Cars parked in the streets or driving on the road are of little interest in an orthophoto or in a GIS. We are demonstrating the ability of automatically detecting nearly all cars shown in an urban aerial photography at a ground sampling distance in the range of 8 cm to 15 cm. The image shapes from those cars can be inpainted to create a car-less orthophoto. In addition, the shapes can also be used to remove the effect of cars on the digital elevation model by filtering the elevations representing cars and thereby improving the Bald Earth DTM. We study the use of redundancy from image overlaps and the introduction of color to further improve the score beyond the current high success.
We propose a framework for a complex visualization environment suitable for archaelogical applications. Given 2D and 3D data derived from appropriate acquisition processes, the scene is organized in a structure that c...
详细信息
ISBN:
(纸本)1581134479
We propose a framework for a complex visualization environment suitable for archaelogical applications. Given 2D and 3D data derived from appropriate acquisition processes, the scene is organized in a structure that can easily be incorporated into a database. Special care is taken on attributes such as time and likelihood of scientific hypothesises which are important for a correct interpretation of the excavation site. After a preprocessing step, the database content can directly be used to visualize the scene in a standalone virtual reality installation in a museum as well as on the internet.
Existing visual surveillance systems typically require that human operators observe video streams from different cameras, which becomes infeasible if the number of observed cameras is ever increasing. In this paper, w...
详细信息
In this paper we consider the limitations of Linear Discriminative Analysis (LDA) when applying it for large-scale problems. Since LDA was originally developed for two-class problems the obtained transformation is sub...
详细信息
ISBN:
(纸本)9789898111692
In this paper we consider the limitations of Linear Discriminative Analysis (LDA) when applying it for large-scale problems. Since LDA was originally developed for two-class problems the obtained transformation is sub-optimal if multiple classes are considered. In fact, the separability between the classes is reduced, which decreases the classification power. To overcome this problem several approaches including weighting strategies and mixture models were proposed. But these approaches are complex and computational expensive. Moreover, they were only tested for a small number of classes. In contrast, our approach allows to handle a huge number of classes showing excellent classification performance at low computational costs. The main idea is to split the original data into multiple sub-sets and to compute a single LDA space for each sub-set. Thus, the separability in the obtained subspaces is increased and the overall classification power is improved. Moreover, since smaller matrices have to be handled the computational complexity is reduced for both, training and classification. These benefits are demonstrated on different publicly available datasets. In particular, we consider the task of object recognition, where we can handle up to 1000 classes.
We present a hybrid multi-line scan approach which enables simultaneous acquisition of light field & photometric stereo data. While light fields capture mostly large-scale surface deviations and rely on visible su...
详细信息
We show how to train a Convolutional Neural Network to assign a canonical orientation to feature points given an image patch centered on the feature point. Our method improves feature point matching upon the state-of-...
详细信息
ISBN:
(纸本)9781467388511
We show how to train a Convolutional Neural Network to assign a canonical orientation to feature points given an image patch centered on the feature point. Our method improves feature point matching upon the state-of-the art and can be used in conjunction with any existing rotation sensitive descriptors. To avoid the tedious and almost impossible task of finding a target orientation to learn, we propose to use Siamese networks which implicitly find the optimal orientations during training. We also propose a new type of activation function for Neural Networks that generalizes the popular ReLU, maxout, and PReLU activation functions. This novel activation performs better for our task. We validate the effectiveness of our method extensively with four existing datasets, including two non-planar datasets, as well as our own dataset. We show that we outperform the state-of-the-art without the need of retraining for each dataset.
暂无评论