Georg Poier

I'm a postdoctoral researcher in the Team of Prof. Horst Bischof. Very roughly, my research interest is in machine learning for computer vision :-). Find more details, including links to projects and publications below.

If you are interested in our work or a student project, feel free to send me an e-mail.


Projects for Students

Find my open student projects at the institutes' "Student Projects"-page by searching for my name.


Selected Projects

Human Shape Estimation
The goal of this work is to develop methods to reconstruct the shape of a person given a single color image.
Such methods provide the basis for applications like human computer interaction, shape analysis or fitting of, e.g., clothes or medical devices.
The figure shows depth image estimates for color images of persons, which can be a first step towards 3D reconstruction.
(Human) Depth estimation samples with input image and output visualizations
Hand Pose Estimation
The goal of this work is to develop methods to estimate the hand pose, i.e., the joint positions, given an image of the hand.
Such methods provide the basis for applications like human computer interaction, activity analysis, and sign language recognition. One of our implementations, e.g., found the way into a product of a partner company, which used the estimated hand poses for interaction with computer games.
Find related publications and further material at poier.github.io.
Hand pose estimation samples with input image and output visualizations
Semantic Segmentation
The goal of this work was to develop methods to segment petroglyphs pecked into the rocks of Valcamonica (northern Italy) thousands of years ago.
The main challenge stemmed from the weathering, which the rocks have undergone in the meantime.
Find related publications and further material of our work at poier.github.io and a general project overview with links to more material and videos at 3d-pitoti.eu.
Segmentation samples with input depth and mask visualization

Publications

A list of my publications including links to additional material can be found at poier.github.io.

DBLP | ResearchGate | PURE | Google Scholar