Computer Vision Group research

Biomedicine

2020-11-09T11:36:05+00:00

Biomedicine Below, you can find a list of publications where we partially or entirely focus on biomedicine. Also many of our other methods can be applied to biomedical data. Contact Related publications

Convex Relaxation Methods

2015-09-25T15:46:37+00:00

Convex Relaxation Methods Contact: Thomas Möllenhoff, Evgeny Strekalovskiy A popular and well established paradigm for modeling computer vision problems is through energy minimization. In practice, almost all functionals providing a realistic model are non-convex and even NP-hard. They are thus hard to solve and a direct minimization usually leads to poor local minima.

Deep Learning

2020-12-01T11:16:13+00:00

deeplearning Deep Learning Deep learning is a powerful machine learning framework that has shown outstanding performance in many fields. The main power of deep learning comes from learning data representations directly from data in a hierarchical layer-based structure.

E-NeRF: Neural Radiance Fields from a Moving Event Camera

2023-03-16T09:29:49+00:00

E-NeRF: Neural Radiance Fields from a Moving Event Camera [ ] Code Code can be found on https://github.com/knelk/enerf. Abstract Estimating neural radiance fields (NeRFs) from ideal images has been extensively studied in the computer vision community. Most approaches assume optimal illumination and slow camera motion. These assumptions are often violated in robotic applications, where images contain motion blur and the scene may not have suitable illumination. This can cause significant pro…

Geometry Processing

2021-01-15T10:56:09+00:00

Geometry Processing Geometry processing is concerned with the acquisition, analysis and manipulation of geometric data. The field is very broad and algorithms aim at improving 3D reconstructions, finding correspondences between objects, shape interpolation, or analysing physical properties of scanned data. Many applications in geometry processing have counterparts in image processing, e.g. object detection, but geometric data poses special challenges in terms of how to formulate mathematical co…

Image-based 3D Reconstruction

2017-05-22T13:06:59+00:00

Image-based 3D Reconstruction Contact: Prof. Dr. Daniel Cremers For a human, it is usually an easy task to get an idea of the 3D structure shown in an image. Due to the loss of one dimension in the projection process, the estimation of the true 3D geometry is difficult and a so called ill-posed problem, because usually infinitely many different 3D surfaces may produce the same set of images.

Image Segmentation

2012-01-20T11:57:22+00:00

Image Segmentation Contact: Claudia Niewenhuis, Maria Klodt Image segmentation aims at partitioning an image into n disjoint regions. Since this problem is highly ambiguous additional information is indispensible. This can be given as user input, e.g. scribbles on the image, additional constraints such as the center of gravity and the major axes of the object or learned from a given database. We formulate mostly convex energy functionals to solve this problem.

research:lsdslam

2015-07-27T22:44:35+00:00

no summary

Marker-less Motion Capture

2012-01-20T11:58:56+00:00

Marker-less Motion Capture In this project, we develop statistical and energy minimization methods to tracking articulated 3D objects from multiple camera views. Such techniques are of central importance in particular for markerless motion capture. The human motion sequence extracted from multiple videos can subsequently be used to animate virtual characters as commonly done in action movies.

MonoRec: Semi-Supervised Dense Reconstruction in Dynamic Environments from a Single Moving Camera

2021-06-15T09:38:16+00:00

MonoRec: Semi-Supervised Dense Reconstruction in Dynamic Environments from a Single Moving Camera Contact: Nan Yang, Lukas Von Stumberg, Niclas Zeller Code: Abstract In this paper, we propose MonoRec, a semi-supervised monocular dense reconstruction architecture that predicts depth maps from a single moving camera in dynamic environments. MonoRec is based on a multi-view stereo setting which encodes the information of multiple consecutive images in…

Nanocopter

2015-05-19T23:22:40+00:00

Nanocopter Contact: Jakob Engel, Jörg Stückler In this research project we will explore visual navigation methods for Nanocopters, extremely small Quadrocopter with a flight-weight of less than 50g. Possible future applications are e.g. exploration of a collapsed building after a natural desaster, surveillance and inspection of difficult-to-reach areas, or simply as personal flying camera, to take pictures from a fully new perspective. In contrast to regular quadrocopters with a flight-weight …

research:omni-lsdslam

2015-09-21T22:09:23+00:00

no summary

Optical Flow Estimation

2020-10-27T13:04:00+00:00

Optical Flow Estimation Estimating the motion of every pixel in a sequence of images is a problem with many applications in computer vision, such as image segmentation, object classification,visual odometry, and driver assistance. In general, optical flow describes a sparse or dense vector field, where a displacement vector is assigned to certain pixel position, that points to where that pixel can be found in another image. In the context of scene flow estimation, which is performed on images …

research:overview

2020-10-15T14:28:03+00:00

no summary

Photometry-Based Reconstruction

2020-10-27T13:15:00+00:00

Photometry-Based Reconstruction We are concerned with the reconstruction of the 3D world based on the interaction between shape, illumination and material. RGB images provide observations from which we can infer the above, by solving an ill-posed inverse rendering problem. This enables reconstructions with high-frequency geometric information and meaningful albedo estimates, allowing for plausible rendering under novel lighting conditions. We are especially interested in Shape-from-Shading an…

Micro Aerial Vehicles (MAVs)

2015-03-14T13:26:41+00:00

Micro Aerial Vehicles (MAVs) Contact: Jörg Stückler, Jakob Engel, Christian Kerl, Vladyslav Usenko In recent years, flying robots such as quadrocopters have gained increased interest in robotics and computer vision research. To navigate safely, these robots need the ability to localize themselves autonomously using their onboard sensors. Potential applications of such systems include the usage as a flying camera, for example to record sport movies or to inspect bridges from the air, as well a…

RGB-D Vision

2017-05-22T11:55:56+00:00

RGB-D Vision Contact: Mariano Jaimez and Robert Maier In the past years, novel camera systems like the Microsoft Kinect or the Asus Xtion sensor that provide both color and dense depth images became readily available. There are great expectations that such systems will lead to a boost of new 3D perception-based applications in the fields of robotics and visual&augmented reality.

Robot Vision

2018-05-01T13:14:07+00:00

Robot Vision Contact: Jörg Stückler, Prof. Dr. Daniel Cremers We develop computer vision methods for robot systems such as micro aerial vehicles and wheeled robots. In recent years, flying robots such as quadrocopters have gained increased interest in robotics and computer vision research. To navigate safely, these robots need the ability to localize themselves autonomously using their onboard sensors.

Scene Flow Estimation

2017-12-06T14:34:56+00:00

Scene Flow Estimation Scene flow is the dense or semi-dense 3D motion field of a scene that moves completely of partially with respect to a camera. The potential applications of scene flow are numerous. In robotics, it can be used for autonomous navigation and/or manipulation in dynamic environments where the motion of the surrounding objects needs to be predicted. Besides, it could complement and improve state-of-the-art Visual Odometry and SLAM algorithms which typically assume to work in rig…

Semi-Dense Monocular Visual Odometry

2014-07-02T08:51:02+00:00

no summary

Shape Analysis

2020-10-06T10:08:07+00:00

Shape Analysis Over the last years, the availability of devices for the acquisition of three-dimensional data like laser-scanners, RGB-D Vision or medical imaging devices has dramatically increased. This brings about the need for efficient algorithms to analyze three-dimensional shapes.

Shape Priors

2012-01-20T11:58:35+00:00

Shape Priors In this project, we introduce into classical image segmentation methods some prior knowledge about which shapes are likely to be in a given image. In particular, we develop metrics on spaces of shapes, statistical models of shape variation and dynamical models which allow to impose a statistical model of the temporal evolution of shape. The respective segmentation processes can cope with large amounts of background clutter, noise and partial or complete occlusions.

3D Reconstruction from a Single Image

2014-02-23T13:47:28+00:00

3D Reconstruction from a single view Contact: Martin Oswald, Eno Toeppe The estimation of 3D geometry from a single image is a special case of image-based 3D reconstruction from several images, but is considerably more difficult since depth cannot be estimated from pixel correspondences. Thus, further prior knowledge or user input is needed in order to recover or infer any depth information.

Stereo LSD-SLAM

2015-07-27T22:55:58+00:00

no summary

Visual SLAM

2020-10-15T15:07:52+00:00

Visual SLAM In Simultaneous Localization And Mapping, we track the pose of the sensor while creating a map of the environment. In particular, our group has a strong focus on direct methods, where, contrary to the classical pipeline of feature extraction and matching, we directly optimize intensity errors.

Projects in Visual SLAM

2020-10-21T00:23:05+00:00

Projects in Visual SLAM