LAVIS (Learning and Visual Systems)

Finished Projects

AMIGO

APOLLON

Virtual reality applications often requires to bake complex light situations for specific objects in a virtual scene into textures, so called lightmaps. These lightmaps can be used in any 3D realtime applications like games or visualization apps. With precalculated lightmaps it is possible to improve the performance and the visual quality in realtime applications. In this project we develop a software tool that supports at least baking of Global Illumination, Ambient Occlusion and Direct Lighting / Shadow Baking.

LOEWE_logo Dieses Projekt (HA-Projekt-Nr. 229/10-05) wurde im Rahmen von Hessen ModellProjekte aus Mitteln der LOEWE – Landes-Offensive zur Entwicklung Wissenschaftlich-ökonomischer Exzellenz, Förderlinie 3: KMU-Verbundprojektvorhaben gefördert.
AUTOMEX

AUTOMEX

Many simulations like crash-tests, pedestrian vehicle interactions, vibration or noise analysis depend on simplified finite element models, which have to be generated from existing 3D-CAD models. The AUTOMEX project develops automated methods to compute medial axis representations for arbitrarily complex thin-walled components such as synthetic materials and cast metal parts. For further information please visit our project website. LOEWE_logo

Dieses Projekt (HA-Projekt-Nr. 300/11-45) wurde im Rahmen von Hessen ModellProjekte aus Mitteln der LOEWE – Landes-Offensive zur Entwicklung Wissenschaftlich-ökonomischer Exzellenz, Förderlinie 3: KMU-Verbundprojektvorhaben gefördert.
docCap

Camera-based Document Capture

Compared to scanners, digital cameras offer a cheap, quick, or even ubiquitious alternative for capturing documents. However, information extraction from the resulting images is hampered by low resolution, noise, and distorsion. These challenges can be tackled by dewarping document images prior to analysis, or by robustifying optical character recognition (OCR) to take low resolution into account.

demo | sample paper (dewarping)
emil

EMIL

Im Rahmen des Forschungsprojekts EMIL wurde eine innovative Möglichkeit geschaffen, Multitouch- und Tangible-User-Interfaces zu erstellen. EMIL kann in der betrieblichen Praxis von Unternehmen genutzt werden, um u.a. Neukunden zu akquirieren und um den Erstellungsprozess solcher Benutzerschnittstellen zu verbessern. Firmen, die mit EMIL erstellte Anwendungen einsetzten, besitzen mit diesen neuartigen Benutzerschnittstellen einen Innovationsvorsprung vor Wettbewerbern. Im Projekt wurde die EMIL-Umgebung erstellt. Diese Umgebung formt einen nutzer-zentrierten Design-Prozess zur Erstellung von Software für interaktive Oberflächen, in dessen Rahmen mit Hilfe der von EMIL bereitgestellten Bibliotheken und Tools interaktive Prototypen erstellt werden. BMBF_logo Mit Hilfe dieser Prototypen können Ersteller solcher Software potenziell neue Aufträge akquirieren, Software näher am Kunden entwickeln und zur Qualitätssicherung der Usability frühzeitig mit dem Entnutzer testen. Die EMIL-Umgebung umfasst unter anderem ein Basis-Framework, ein Autoren-Werkzeug, zwei Applikations-Frameworks und ein System zur Bereitstellung und Sammlung von Design-Wissen.
FIBEVID

FIBEVID

Key to preserving (often endangered) fish species is a cheap, non-invasive sampling strategy to tackle issues such as population estimation and species association tracking. Current manual sampling methods are expensive, of limited coverage, and come with fish injury and mortality. Therefore, the goal of FIBEVID is to employ underwater computer vision technology for an automated surveying of fish. This includes the construction and deployment of suitable underwater camera rigs, as well as the development of semi-automatic localization, tracking, measuring and classification software for underwater species. DAAD_logo FIBEVID is conducted in cooperation with the National University of Sciences and Technology (NUST) in Islamabad / Pakistan. The project is sponsored by the German Academic Exchange Service (DAAD).
KARBON

KARBON

Für die Unterstützung von Planungsvorgängen bei der Bauorganisation setzt kARbon Post-WIMP Technologien ein, um Werkzeuge anzubieten, mit denen planende und ausführende Kräfte kollaborativ zusammenarbeiten. Endgeräte wie Tablets bieten dabei eine Plattform für intuitive, leichtgängige Interaktion auf Basis Deiktischer Gesten (Zeigegesten). Ziel von kARbon ist es, eine verteilte Mehrbenutzerkonferenz zu ermöglichen, bei denen sich am Bau beteiligte Personen wie Architekt und Bauleiter über verschiedene Endgeräte von verschiedenen Orten aus virtuell treffen und gemeinsam Arbeitsschritte planen. Durch solche Rücksprachen als erweiterte Form einer Telefonkonferenz wird die Notwendigkeit von realen Treffen reduziert, was besonders bei Projekten mit weit entfernten Partnern von Vorteil ist. FFDP_logo Dies verringert zum einen den Kostenaufwand derartiger Treffen, zum anderen können die Situation vor Ort und die durchzuführenden Schritte deutlich zielgerichteter, als bei einem reinen Telefongespräch besprochen werden. Daher bieten die kARbon-Werkzeuge ein Mittel zur Reduzierung von Missverständnissen, die sonst zu sehr kostenintensiven Fehlern führen können.
Kephalos

Kephalos

Orthodontic motivated radiographs substantial contributes to radiation exposure of children and adolescents. Especially since the introduction of dental digital volume tomography (DVT), the use of 3D radiographs is increasingly propagated. Because DVT devices typically generate 3D information from several hundred 2D radiographs, the effective dose of a DVT recording is at least an order of magnitude greater than that of a single digital lateral cephalogram. However, for many diagnostics a complete 3D model often is not needed, but only the outer (facial) surface of the facial skeleton. BMBF_logo The aim of Kephalos is to develope a new technique for calculating the exact facial surface of the facial bones based on a) a single lateral cephalogram, b) an optical face scan and c) a statistical model of the correlation between the facial surface and the facial bones.
Kephalos is funded by the Federal Ministry of Education and Research.
KOLIBRI

KOLIBRI

The viability of small and cheap autonomous aircrafts opens up a whole new field of civilian applications. In this research project we are developing a low budget system to do research in the field the swarm behavior of small autonomous aircrafts. Our system is based on an affordable high speed optical tracking system with high accuracy and a software component for remote controlling arial vehicles, both systems are developed in our lab. Based on very small multicopters with a maximum weight of 50g, we investigate autonomous aircraft swarms and develop new algorithms to enable collaborative strategies for solving problems such as generating maps of an environment or identifying best paths to victims. The latest results can be found here.
LowDoseDVT

LowDoseDVT

The aim of this research project is to develop algorithms that allow to calculate high quality three-dimensional reconstruction from cone beam computed tomography data needing fewer projection images than before. In this way the radiation dose for patients can be reduced significantly. In this project we follow three strategies in order to reduce the number of projection images needed. BMBF_logo First, we carefully calibrate the geometric parameters of a cone beam computed tomography machine. Second, we reduce reconstruction artifacts by suitable processing the input images. And third, we develop new regulized algebraic reconstruction algorithms. For further information please visit our project website.
MILAN

MILAN

Innerhalb des Forschungsprojekts MILAN wird die Infrastruktur ICARUS (Infrastructure for Compact Aerial Robots Under Supervision) des CVMR-Labors so zu erweitern, dass die darin verwendeten Multirotoren das Ausführen aggressiver Manöver erlernen. Diese dienen zum Beispiel dem schnellen kontrollierten Ausweichen in Gefahrensituationen, dem Durchführen von Mehrfach-Salti oder dem Minimieren der Flugzeiten beim Verfolgen vorgegebener Trajektorien. Der Fokus liegt dabei auf maschinellen Lernverfahren zur Verbesserung der notwendigen Bewegungsabläufe basierend auf Sensordaten und einem dynamischen Bewegungsmodell. ICARUS entstand im Rahmen der zuletzt geförderten Projekte KOLIBRI und ESCAPE, die in beiden vorangegangenen Jahren auf der CeBIT vorgestellt wurden und jeweils großes Interesse in den Medien fanden.
MTM

MTM Auto Code Generator

Methods-Time Measurement (MTM) is often applied in industrial settings whenever any manual operation or task is performed during the manufacturing process. It is used to analyze and optimize motion sequences of workers and thereby determine standard times in which a worker can complete the individual tasks. Thus one of the basic requirements of MTM is the ability to capture, and classify the movement of workers while executing these tasks. This research project aims to automate this by applying and fusing motion tracking methods based on various sensors. zim_bmwi_logo We develop a system to accurately capture the motion of a worker, employed tools and components and simultaneously segment and classify these motion sequences. For tracking components and tools we investigate a novel optical passive tracking algorithm since it is impractical to augment these parts with any sensor or marker attachment in this setting.
mifor

Multimedia Forensics

The massive growth of multimedia data poses challenges to law enforcement, who are confronted with analysing vast amounts of evidence in digital image and video form. Scenarios are the investigation of public desasters, or of cases of child sexual abuse. Here, multimedia analysis can support forensics by content classification, similarity search, or content clustering.

iCOP project | sample paper (similarity search)
Kephalos

OPTIGAMA

The research project OPTIGAMA (Optical Short Game Motion Analyzer) is embedded in the research project SHOGAMA (Short Game Motion Analyzer) which is a joint project together with the Science & Motion Sports GmbH in Flörsheim and the institute of computer science of the Johannes Gutenberg University Mainz. The aim of SHOGAMA is to develop a highly accurate realtime system for three-dimensional tracking of the golf cub during short game. The complete system will be composed of an inertial measurement unit (IMU) attached to the golfs cub handle as well as a camera observing the head of the golf cub. The aim of OPTIGAMA is to determine the golf cub position and orientation (pose) while it is inside the cameras field of view based on the camera footage. ZIM_BMWI_logo The determined pose can be additionally refined and verified based on the data obtained by the IMU. On the other hand pose information can be achieved only based on IMU data if the golf cub is outside the cameras field of view. This combination allows to monitor the complete golf stroke mechanics.
Optishot

Optishot

Professionals who carry firearms on man (e.g. police officers), must be specially trained to deal with them. The training largely consists of repetition exercises in which the movement, including firing be rehearsed "process-safe". Existing measurement methods are very limited for a detailed analysis of the training shots and practical instructions to improve the sequence of movements. ZIM_BMWI_logo In addition, there is a disease (that is also known by golfer and called "yips"), which may lead to occupational disability, if it is not diagnosed early. In this research project we develop a new measuring method which precisely determine and evaluate the 3D movement and automatically creates instructions to improve the motion sequence.
SHOGAMA

SHOGAMA

Im Rahmen des FuE-Projekts wird ein neuartiges System entstehen, mit dem die Bewegung eines Golfschlägers beim sogenannten Kurzspiel (Putten und Chippen) räumlich und hochpräzise erfasst werden kann. Dabei wird ein optisches, monokulares, passives Trackingverfahren in Kombination mit einer intertialen Messeinheit (englisch intertial measurement unit, IMU) entwickelt werden. Die beiden Sensorquellen gleichen hierbei jeweils ihre gegenseitigen Schwächen aus, um so den Messraum und die Messgenauigkeit des Gesamtsystems zu vergrößern bzw. zu erhöhen. Die Kamera wird so positioniert, dass sie die Schlägerbewegungen nahe des Balls, inklusive des Abschlags, mit maximaler Präzision erfassen kann. Sobald der Schläger den Sichtbereich der Kamera verlässt, wird die Bewegung mit der IMU anhand eines physikalischen Schlagmodells rekonstruiert. Der zeitliche Drift der den IMU Messdaten innewohnt, wird durch die absoluten Messungen der Kamera korregiert, sobald der Schläger wieder in deren Sichtbereich eintritt. Die Algorithmen werden auf einem eingebetteten System implementiert werden, welches dem Golfer innerhalb weniger Sekunden Feedback in Form einer 3D Visualisierung des Schlages liefert. ZIM_BMWI_logo
Die Hochschule RheinMain bearbeitet hierbei im Teilprojekt OPTIGAMA (Optical Short Game Motion Analyzer) im wesentlichen Fragestellungen im Bereich des optischen Trackings von Golfschlägern.
SMULGRAS

SMULGRAS

Content-based image search and 3D modelling are trending fields, with applications from copyright protection over mobile object search and 3D printing to community-based modeling. Key to exploiting large collections of images and 3D models, however, are efficient search and organization mechanisms. Here, SMULGRAS' goal is to investigate a multi-codal search between 2D images and 3D models: Given a camera snapshot, can we find the matching 3D model in a database? And given a 3D model, can we find images showing the corresponding object? FFDP_logo

SMULGRAS is funded by the program Forschung für die Praxis and is executed in collaboration with Prof. Yvonne Jung from Hochschule Fulda and IntelligentGraphics GmbH.

web demo (request access per e-mail)
TECAN

TECAN Bench

TECAN und HSRM erschließen in einer wissenschaftlichen Zusammenarbeit Synergien im Bereich optischer Überwachungsverfahren in der Laborautomation. Die Kooperation ermöglicht insbesondere dem wissenschaftlichen Nachwuchs der HSRM, an aktuellen und praxisrelevanten Fragestellungen zu arbeiten. Gegenstand des Projekts ist eines gemeinsames interdisziplinärer Projekt im Bereich der optischen Worktable-Erkennung.
TRACKSHOT

TRACKSHOT

The research project TRACKSHOT (Tracking of Shotgun and Target for Olympic Trap and Skeet) is a joint research project with the German Shooting Federation (DSB).The aim of the project is to detect and track the position and orientation (pose) of the shotgun together with the position of the target simultaneously in a common coordinate system. The resulting performance-diagnostic hardware component and a suitable software application should enable both the athlete and the trainer to have an accurate and timely evaluation and interpretation of the data. Beside the development of a robust and accurate marker-based tracking system, one of the main focuses of the project is to set the basis for a completely marker-less system, that enables tracking without the need of hardware to be attached to the sports equipment in order to have a completely non-reactive system. BISP_logo Particularly suitable for this initial tests for the development of such a marker less optical tracking system is the shotgun due to its size and thus good visibility. Thus, a feasibility study for marker-less tracking is conducted within the project. If this proves to be feasible, it would be dealt with in follow-up applications for other sports equipment such as air gun or bow.
VEDUS

VEDUS

Many tasks in the field of urban planning, such as noise protection measures, the incidence of light, or radio coverage require 3D data. Complete 3D city models thus gradually replace conventional maps. But the creation and maintenance of 3D city models is very complex and costly. The goal of this research project is to develop a system that allow editing of 3D city models by various users using al kind of different devices. Users should be placed in a position to visualize and collaboratively edit the models on all kind of devices from the traditional desktop PC to Tablet PCs and smartphones. LOEWE_logo

Dieses Projekt (HA-Projekt-Nr. 315/12-05) wurde im Rahmen von Hessen ModellProjekte aus Mitteln der LOEWE – Landes-Offensive zur Entwicklung Wissenschaftlich-ökonomischer Exzellenz, Förderlinie 3: KMU-Verbundprojektvorhaben gefördert.
TIT

Visual Learning from the Web

Web-based portals like Flickr or YouTube offer huge amounts of images and videos, combined with rich context (such as tags, descriptions, comments, ratings, etc). This information can be exploited for a multi-modal auto-understanding of web-based multimedia. By training visual recognition systems at larger scale and combining their output with other signals, smarter search and recommendation can be facilitated. Core research issues are learning from unreliable labels, domain drift, and new forms of supervision to learning.

MOONVID project | demo | sample paper (demographics prediction)