Filetype
We present a model for generating 3d articulated pedestrian locomotion in urban scenarios, with synthesis capabilities informed by the 3d scene semantics and geometry. We reformulate pedestrian trajectory forecasting as a structured reinforcement learning (RL) problem. This allows us to naturally combine prior knowledge on collision avoidance, 3d human motion capture and the motion of pedestrians
Precipitation over the Tibetan Plateau (TP) has major societal impacts in South and East Asia, but its spatiotemporal variations are not well understood, mainly because of the sparsely distributed in situ observation sites. With the help of the Global Precipitation Measurement satellite product IMERG and the ERA5 dataset, distinct precipitation seasonality features over the TP were objectively cla
We derive an expression for the error introduced by the second-order accurate temporal finitedifference (FD) operator, as present in the FD, pseudospectral and spectral element methods for seismic wave modeling applied to time-invariant media. The 'time-dispersion' error speeds up the signal as a function of frequency and time step only. Time dispersion is thus independent of the propagation path,
This thesis explores methods for estimating 3D models using depth sensors andfinding low-rank approximations of matrices. In the first part we focus on how toestimate the movement of a depth camera and creating a 3D model of the scene.Given an accurate estimation of the camera position, we can produce dense 3Dmodels using the images obtained from the camera. We present algorithms thatare both accu
Traditional matrix factorization methods approximate high dimensional data with a low dimensional subspace. This imposes constraints on the matrix elements which allow for estimation of missing entries. A lower rank provides stronger constraints and makes estimation of the missing entries less ambiguous at the cost of measurement fit. In this paper we propose a new factorization model that further
Are we using the right potential functions in the Conditional Random Field models that are popular in the Vision community? Semantic segmentation and other pixel-level labelling tasks have made significant progress recently due to the deep learning paradigm. However, most state-of-the-art structured prediction methods also include a random field model with a hand-crafted Gaussian potential to mode
Human body pose and shape estimation is an important and challenging task in computer vision. This paper presents a novel method for estimating 3D human body pose and shape from several RGB images, using detected joint positions in the images and based on a parametric human body model. Firstly, the 2D joint points of the RGB images are estimated using a deep neural network, which provides a strong
With the advent of cheap sensors and computing capabilities as well as better algorithms it is now possible to do structure from motion using crowd sourced data. Individual estimates of a map can be obtained using structure from motion (SfM) or simultaneous localization and mapping (SLAM) using e.g. images, sound or radio. However the problem of map merging as used for collaborative SLAM needs fur
In this paper we introduce a fast and robust structure-less alternative to full bundle adjustment. The method is based on optimizing algebraic errors for trilinear constraints from triplets of views. It is shown that the error generated by a triplet of views can be described by a fixed triangular matrix regardless of the number of feature correspondences between the views. The method has been eval
In recent years, great progress has been made on the problem of 3D scene reconstruction using depth sensors. On a large scale, these reconstructions look impressive, but often many fine details are lacking due to limitations in the sensor resolution. In this paper we combine two well-known principles for recovery of 3D models, namely fusion of depth images with photometric stereo to enhance the de
Multiple rotation averaging is an important problem in computer vision. The problem is challenging because of the nonlinear constraints required to represent the set of rotations. To our knowledge no one has proposed any globally optimal solution for the case of simultaneous updates of the rotations. In this paper we propose a simple procedure based on Lagrangian duality that can be used to verify
Due to advances in technology the amount of images from portable cameras has increased tremendously in recent years. Nowadays, most new mobile phones and cars have multiple cameras. Drones and other robots are often equipped with cameras as well. To be able to capture an image is not the same as to understand the content of it. Computer vision deals with the task of giving machines the ability to
A large number of medical examinations involve images in some way. Images can be used for diagnostics, follow-up studies and treatment planning. In this thesis mathematical methods have been developed and adapted in order to analyze medical images. Several applications for different imaging modalities have been studied and the usefulness of such methods is demonstrated.A complete system for detect
The field of high dynamic range imaging addresses the problem of capturing and displaying the large range of luminance levels found in the world, using devices with limited dynamic range. In this paper we present a novel tone mapping algorithm that is based on K-means clustering. Using dynamic programming we are able to not only solve the clustering problem efficiently, but also find the global op
Localization is an important enabling technology for many applications, such as wireless sensor networks, emergency rescue services, civil defense and transportation. Suppose that a room is equipped with several microphones (or sensors), and one person is making a sound while moving around in the room. Can one find microphone and sound source positions as well as reconstruct a room geometry? The a
The central problem of this thesis is locating several sources and simultaneously locating the positions of the sensors. The measurements captured by the sensors are time of arrival (TOA), time difference of arrival (TDOA), unsynchronized TDOA, or received signal strength indication (RSSI), all a variation of distance measurement between sensors and sources. Signals can be either sound or radio fo
The problem of estimating receiver-sender node positionsfrom measured receiver-sender distances is a key issue indifferent applications such as microphone array calibration, radioantenna array calibration, mapping and positioning using ultrawidebandand mapping and positioning using round-trip-timemeasurements between mobile phones and Wi-Fi-units. Thanks torecent research in this area we have an i
Indoor positioning for smartphone usershas received a lot of attention in recent years. Whilemany solutions have been developed, most rely on aneed for pre-deployment of infrastructure or collectingground truth data to train on. In this paper we see whatcan be done using existing WiFi-infrastructure andReceived Signal Strength from these to smartphones,not using any calibration of the signal envir
Precipitation in clouds can form by either warm-rain or ice crystal processes, referred to as warm and cold formation pathways, respectively. Here, we investigate the warm and cold pathway contributions to surface precipitation in simulated continental convective storms. We analyze three contrasting convective storms that are cold-based, slightly warm-based and very warm-based. We apply tracer-tag
