Research
My research is on machine learning and computer vision for video processing. Here are my recent
projects and papers:
|
|
Synthehicle: Multi-Vehicle Multi-Camera Tracking in Virtual Cities
F. Herzog,
J. Chen,
T. Teepe,
J. Gilg,
S. Hörmann,
and G. Rigoll
arXiv, 2022
arXiv /
code /
project website
We propose a massive new dataset for multi-camera vehicle tracking in non-overlapping and
overlapping camera views with full 2D and 3D annotations and instance, semantic, and panoptic
segmentation ground truth.
Check out our project website to
download the new dataset!
|
|
Towards a Deeper Understanding of Skeleton-based Gait Recognition
T. Teepe,
J. Gilg,
F. Herzog,
S. Hörmann,
and G. Rigoll
arXiv, 2022
arXiv /
code
***Accepted for the 17th IEEE Computer Society Workshop on
Biometrics 2022***
We propose an approach based on Graph Convolutional Networks (GCNs) that combines higher-order
inputs, and residual networks to an efficient architecture for gait recognition.
|
|
The Box Size Confidence Bias Harms Your Object Detector
J. Gilg,
T. Teepe,
F. Herzog,
and G. Rigoll
arXiv, 2022
arXiv /
code
***Accepted for WACV 2023***
We demonstrate how to modify the histogram binning calibration to improve performance through
conditional confidence calibration.
|
|
Face Aggregation Network For Video Face Recognition
S. Hörmann,
Z. Chao,
M. Knoche,
F. Herzog,
and G. Rigoll
IEEE International Conference on Image Processing (ICIP), 2021
IEEE Article
***Accepted for ICIP 2021***
We suggest a permutation invariant U-Net architecture capable of processing an arbitrary number of
frames in video face recognition.
|
|
GaitGraph: Graph Convolutional Network for Skeleton-Based Gait Recognition
T. Teepe,
A. Khan,
J. Gilg,
F. Herzog,
S. Hörmann,
and G. Rigoll
IEEE International Conference on Image Processing (ICIP), 2021
arXiv /
code
***Accepted for ICIP 2021***
A novel Graph Convolutional Network (GCN) to obtain a modern model-based approach for gait
recognition.
|
|
Lightweight Multi-Branch Network for Person Re-Identification
F. Herzog,
X. Ji,
T. Teepe,
S. Hörmann,
J. Gilg, and
G. Rigoll
IEEE International Conference on Image Processing (ICIP), 2021
arXiv /
code
***Accepted for ICIP 2021***
Lightweight network that combines global, part-based, and channel features in a unified multi-branch
architecture.
|
|
Dissected 3D CNNs: Temporal Skip Connections for Efficient Online Video Processing
O. Köpüklü,
S. Hörmann,
F. Herzog,
H. Cevikalp, and
G. Rigoll
arXiv, 2020
arXiv
/
code
***Accepted for Computer Vision and Image Understanding
Journal***
We propose dissected 3D CNNs to address several serious handicaps of 3D ResNet models during online
operation.
|
|
Comparative Analysis of CNN-Based Spatiotemporal Reasoning in Videos
O. Köpüklü,
F. Herzog, and
G. Rigoll
ICPR International Workshops and Challenges. ICPR 2021.
arXiv
/
code
/
video
/
publication
***Accepted for ICPR International Workshops 2021***
A comparative analysis of different spatiotemporal modeling techniques for action and gesture
recognition.
|
|
sparselandtools: A Python package for sparse
representations and dictionary learning, including
matching pursuit, K-SVD and applications
Python package via Zenodo
(10.5281/zenodo.4916395), 2021
Sparselandtools is a Python 3 package that provides implementations for sparse representations and
dictionary learning.
|
|