Filetype
Understanding the loss surface of neural networks is essential for the design of models with predictable performance and their success in applications. Experimental results suggest that sufficiently deep and wide neural networks are not negatively impacted by suboptimal local minima. Despite recent progress, the reason for this outcome is not fully understood. Could deep networks have very few, if
Robust fitting of geometric models is a core problem in computer vision. The most common approach is to use a hypothesize-and-test framework, such as RANSAC. In these frameworks the model is estimated from as few measurements as possible, which minimizes the risk of selecting corrupted measurements. These estimation problems are called minimal problems, and they can often be formulated as systems
A fully automatic system for abdominal organ segmentation is presented. As a first step, an organ localization is obtained via a robust and efficient feature registration method where the center of the organ is estimated together with a region of interest surrounding the center. Then, a convolutional neural network performing voxelwise classification is applied. Two convolutional neural networks o
Despite of having no explicit shape model, multi-atlas approaches to image segmentation have proved to be a top-performer for several diverse datasets and imaging modalities. In this paper, we show how one can directly incorporate shape regularization into the multi-atlas framework. Unlike traditional multi-atlas methods, our proposed approach does not rely on label fusion on the voxel level. Inst
Length and area regularization are commonplace for inverse problems today. It has however turned out to be much more difficult to incorporate a curvature prior. In this paper we propose several improvements to a recently proposed framework based on global optimization. We identify and solve an issue with extraneous arcs in the original formulation by introducing region consistency constraints. The
We propose new feature-based methods for rigid and affine image registration. These are compared to state-of-the-art intensity-based techniques as well as existing feature-based methods. On challenging datasets of brain MR and whole-body CT images, a significant improvement in terms of speed, robustness to outlier structures and dependence on initialization is shown.
The deep sea is the largest habitat on earth. Its three great faunal environments - the twilight mesopelagic zone, the dark bathypelagic zone and the vast flat expanses of the benthic habitat- are home to a rich fauna of vertebrates and invertebrates. In the mesopelagic zone (150-1000 in), the down-welling daylight creates an extended scene that becomes increasingly dimmer and bluer with depth. Th
Description: Python® is a free open-source language and environment that has great potential in scientific computing. Computing with Python presents the programming language in close connection wit mathematical applications. The approach of the book is concept based rather than a systematic introduction to the language. It is written for a mathematical readership and is aimed at students with a ma
In this paper we consider a mobile platform performing partial hand-eye calibration and Simultaneous Localisation and Mapping (SLAM) using images of the floor along with the assumptions of planar motion and constant internal camera parameters. The method used is based on a direct parametrisation of the camera motion, combined with an iterative scheme for determining the motion parameters from inte
Structure from sound can be phrased as the problem of determining the position of a number of microphones and a number of sound sources given only the recorded sounds. In this paper we study minimal structure from sound problems in both TOA (time of arrival) and TDOA (time difference of arrival) settings with collinear constraints on e.g. the microphone positions. Three such minimal cases are anal
In this paper we suggest an improvement to a recent algorithm for estimating the pose and ego-motion of a camera which is constrained to planar motion at a constant height above the floor, with a constant tilt. Such motion is common in robotics applications where a camera is mounted onto a mobile platform and directed towards the floor. Due to the planar nature of the scene, images taken with such
Standard structure tensors provide a robust way of directionality estimation of waves (or edges) but only for the case when they do not intersect. In this work, a structure tensor extension using a one-way wave equation is proposed as a tool for estimating directionality in seismic data and images in the presence of conflicting dips. Detection of two intersecting waves is possible in a two-dimensi
In this paper the feasibility to extract the proportion of pigs located in different areas of a pig pen by advanced image analysis technique is explored and discussed for possible applications. For example, pigs generally locate themselves in the wet dunging area at high ambient temperatures in order to avoid heat stress, as wetting the body surface is the major path to dissipate the heat by evapo
Second derivative regularization methods for dense stereo matching is a topic of intense research. Some of the most successful recent methods employ so called binary fusion moves where the combination of two proposal solutions is computed. In many cases the fusion move can be solved optimally, but the approach is limited to fusing pairs of proposals in each move. For multiple proposals iterative b
We propose a layered statistical model for image segmentation and labeling obtained by combining independently extracted, possibly overlapping sets of figure-ground (FG) segmentations. The process of constructing consistent image segmentations, called tilings, is cast as optimization over sets of maximal cliques sampled from a graph connecting all non-overlapping figure-ground segment hypotheses.
We present a system for structure from motion estimation using additional positioning data such as GPS data. The system incorporates the additional data in the outlier detection, the initial estimates and the final bundle adjustment. The initial solution is based on a novel objective function which is solved using convex optimization. This objective function is also used for outlier detection and
We introduce a new dataset, Human3.6M, of 3.6 Million 3D Human poses, acquired by recording the performance of 11 subjects, under 4 different viewpoints, for training realistic human sensing systems and for evaluating the next generation of human pose estimation models. Besides increasing the size the current state of the art datasets by several orders of magnitude, we aim to complement such datas
A one-dimensional continuum-mechanical model of axonal elongation due to assembly of tubulin dimers in the growth cone is presented. The conservation of mass leads to a coupled system of three differential equations. A partial differential equation models the dynamic and the spatial behaviour of the concentration of tubulin that is transported along the axon from the soma to the growth cone. Two o
The problem of finding a low rank approximation of a given measurement matrix is of key interest in computer vision. If all the elements of the measurement matrix are available, the problem can be solved using factorization. However, in the case of missing data no satisfactory solution exists. Recent approaches replace the rank term with the weaker (but convex) nuclear norm. In this paper we show
