Status: taken Supervisor: Sebastian Zambanini Problem Statement Shadows cast in remote sensing images complicate their analysis due to the low contrast in the poorly illuminated image regions. Shadow removal/compensation methods aim at automatically reducing shadow effects, either by a physical or learning-based  model. Goal The goal of this work is to explore and examine methods … Continue reading Shadow Removal in Satellite Images →

Status: taken Supervisor: Sebastian Zambanini Problem Statement Photo triage is the process of grouping similar photos of a large collection and finally selecting the most preferred one in each group. Especially the selection of “the” best photo of a group is challenging as the relevant image features are not obvious to identify and heavily scene-dependent. … Continue reading Automatic Triage of Photo Collections →

Supervisor: Fabian Hollaus Description In the DI4DH project we are developing the DocScan app, which is dedicated to the imaging of documents. Your task is to refine the app, or to implement new tasks. Tasks that can be implemented are for example: Camera control using CameraX Save settings using Room persistence library Workflow Definition of … Continue reading Mobile app development →

Status: open Supervisor: Markus Diem, Florian Kleber, Christina Bartosch (University of Vienna) Exhibition catalogues comprise information containing accompanying texts, name of the paintings, owner of the paintings, etc. An example image shown on this page illustrates a page of an catalogue comprising consecutive numbers, name of the paintings and owners, and the hall number. The … Continue reading Layout Analysis and OCR of art exhibition catalogues →

Supervision Michael Reiter, Roxane Licandro Problem Statement Acute Leukaemia is a disease induce by genetic alterations of blood progenitor cells, which influences the blood generation process (haematopoiesis) and leads to the proliferation of undifferentiated (leukaemic) cells. Flow CytoMetry (FCM) measurements are used to reliably assess these cancer cells (blasts) and to quantify their ratio to … Continue reading Normalized Flowcytometry Representation of Blood Cell Populations of Children affected by Acute Leukaemia →

Status: available Supervisor: Michael Reiter, Christopher Pramerdorfer The purpose of this work is to classify sequences of dermatoscopy images as “changing” or “not changing”. The term “change” refers to a depicted mole or skin lesion, and a given sequence should be classified as “changing” if the depicted mole changes in a form that indicates  a tumor, not simply … Continue reading Dermoscopic Image Analysis by Deep Learning →

Status: available Supervisor: Martin Kampel Problem Statement Automatically detecting the gender of a person or estimating his/her age is valuable in many fields of work. In the area of computer vision exist a variety of possible approaches, mostly based on machine learning, to accomplish this task. A higher complexity arises due to the claim to … Continue reading Deep-Learning for Gender Classification/Age Estimation →

Status: open Supervisor: Markus Diem, Florian Kleber, Stefan Fiel Document layout analysis deals with the layout structure of document images, thus segmenting a page into homogeneous image regions. Within the project READ a framework for layout analysis is currently developed. The layout analysis allows to detect text regions (text lines, text blocks, etc.). The main … Continue reading Document Layout Analysis for Newspapers →

Status: open Supervisor: Sebastian Zambanini Problem Statement In the DeVisOr project, historical aerial images are registered to modern satellite images for the purpose of geo-referencing. However, feature based registrations only deliver coarse registration that needs to be refined for improved accuracy. Goal The goal of this work is to investigate the application of fine registration … Continue reading Fine Registration of Historical Aerial Images and Present-Day Satellite Images →

Status: available Supervisor: Sebastian Zambanini Problem Statement Learning local image descriptors by means of deep convolutional neural nets [1,2] has recently shown to produce stronger features than traditional hand-crafted ones such as SIFT [3]. However, these nets have been trained and evaluated on general scenarios of (wide-basline) object matching. For the DeVisOr project, matching historical … Continue reading Image Descriptor Learning for Matching Historical Aerial Images with Present-Day Satellite Images →