We span a very broad field of diverse research topics focused around visual surveillance. Below you can find a few of our research highlights.
Our research ranges from specific detection tasks (e.g. pedestrians or vehicles) to generic object classes (e.g. obstacle avoidance).
|Visual Object Tracking
We tackle both single and multiple object tracking to enable automated video analysis and behavior understanding.
We address face detection, recognition and tracking to analyse facial expressions and gestures.
The building block for many vision research fields, including person re-identification, food recognition or fine-grained classification.
Real-time hand pose estimation and gesture analysis to support human computer interaction.
Human activity recognition for sports game analysis and unusual event detection.
We address semantic segmentation for archaeological and land surveying applications based on 3D data and multi-spectral aerial images.
Several industrial research projects/contract research.
Our research activities range from object detection and tracking to obstacle detection, from fine-grained classification to semantic segmentation, and from pose estimation to superresolution.
Embedded Machine Learning (start 2019, Christian Doppler Laboratory)
The CD Lab will conduct embedded machine learning research with the following target applications: Object identification in still and moving images, and scene analysis and recognition.
Click here for further details.
DigiColl (start 2020, FFG Smart Cities)
The aim of the project is to introduce a nationwide smart digital waste collection system in order to have a lasting effect on the separation behaviour and to optimise collection routes. An essential aspect of this is the transition from analogue waste collection to a smart service for citizens, who will be integrated bidirectionally into the waste disposal processes.
KI-Waste (start 2021, Green Tech 100)
Climate and environmental protection has top priority in Styria. In order to make a sustainable contribution, modern processes in an innovative circular economy are required. In the KI-Waste project, the recycling process in waste management is to be fundamentally and sustainably modernized. Using artificial intelligence, data generated by image recognition in waste processing is jointly analyzed with time series data in order to re-model and optimize the overall process in the recycling machines.
INTERACT (start 2020, FFG Mobilität der Zukunft)
Improved holistic assessment of pedestrian protection.
The aim of INTERACT is to improve the effectivity assessment of active pedestrian safety systems, which contributes to the prevention of pedestrian accidents and the reduction of injury severities.
HighScene (start 2021, FFG Mobilität der Zukunft)
Online scenario generation from motorway video infrastructure.
The goals of High-Scene are, on the one hand, to increase road safety (reduce the number of accidents) and, on the other hand, to improve and increase the realism of simulations of automated driving.
iLIDS4SAM (start 2020, FFG IKT der Zukunft)
Integrated LiDAR sensors for safe & smart automated mobility.
Details on separate project page, click here.
Instant Avatar (start 2020)
Instant generation of photorealistic human models with mobile devices.
Dynamic Ground Truth (start 2017, FFG Mobilität der Zukunft)
Focus on semantic segmentation to enable efficient testing of Advanced Driver Assistance Systems (ADAS) and Automated Driving Functions (ADF).
This project is funded by the Austrian Ministry for Transport, Innovation and Technology and coordinated by AVL List Gmbh.
Automated Pedestrian Traffic Light (start 2021)
Model and hardware improvements for the next generation traffic lights which use object detection, tracking and intent prediction to optimize pedestrian traffic flow and automatically control traffic lights.
Several industrial research projects/contract research.
The applications range from activity recognition, to OCR, object recognition, detection and tracking.
DARKNET (2016-2018, FFG KIRAS)
Focus on text recognition and generative models.
3D Toll (2016-2018, FFG Bridge)
Focus on 3D reconstruction from image sequences and object detection for toll systems.
Project partner: EFKON AG.
Automated Traffic Light (2015-2018)
Object detection, tracking and intent prediction to optimize pedestrian traffic flow and automatically control traffic lights.
DIANGO (2013-2016, FFG KIRAS)
Focus on co-training from heterogeneous sources and image classification.
MANGO (2013-2016, FFG Bridge)
Focus on food recognition, image representation and fine-grained classification on distributed systems.
3D-PITOTI (2013-2016, EU FP7)
Focus on segmentation of 3D scans of ancient art carved into rocks.
Advanced Learning for Tracking and Detection in Medical Workflow Analysis (2011-2015, FWF DACH)
Focus on online learning for multiple object detection and tracking in a multi-camera environment.
FACTS (2011-2014, FFG RSA)
Focus on face detection, tracking and recognition, as well as emotion recognition.
Quality Eye (2013, FFG Innovation Check+)
Focus on food recognition and segmentation.
Project partner: JIPP.IT.
Kiwi Cloud & Kiwi Background (2012-2013, both FFG)
Focus on online learning and background modeling.
Project partner: Kiwi Security.
SHARE (2011-2013, FFG I2Vsplus)
Focus on online learning for person counting on embedded devices.
Mobi-Trick (2010-2013, FFG FIT-IT)
Focus on online learning for car detection and tracking using embedded devices.
OUTLIER (2009-2012, FFG FIT-IT)
Focus on image representations for unusual event recognition.
MDL (2009-2012, FFG KIRAS)
Focus on face recognition, detection and tracking using multiple instance learning, semi-supervised learning and online learning.
ECV (2008-2012, FFG COMET, contract research)
Focus on person re-identification.
SECRET (2009-2011, FFG KIRAS)
Focus on activity recognition.
AUTOVISTA (2007-2010, FFG FIT-IT)
Focus on person detection and tracking.
Highway Monitoring (2008-2009, FFG FIT-IT)
Focus on car detection and tracking.
VITUS-2 (2005-2006, FFG)
Focus on car detection and tracking for tunnel safety.
FSP/JRP Cognitive Vision (2004-2006, FWF)
Focus on online learning and object detection.