Mateusz Kozinski

Mateusz Kozinski

Auto-generated profile, based on publicly available information

Mateusz Kozinski

Institute for Computer Graphics and Vision Graz University of Technology Austria

Research Focus & Metrics

Academic Career Length

15Years (2012-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 June 2018
Verified
Conference Paper
Closed access
AMAgata Mosinska
Agata Mosinska
PMPablo Marquez-Neila
Pablo Marquez-Neila
MKMateusz Kozinski
Mateusz Kozinski
PFPascal Fua
Pascal Fua
Delineation of curvilinear structures is an important problem in Computer Vision with multiple practical applications. With the advent of Deep Learning, many current approaches on automatic delineation have focused on finding more powerful deep architectures, but have continued using the habitual pixel-wise losses such as binary cross-entropy. In this paper we claim that pixel-wise losses alone are unsuitable for this problem because of their inability to reflect the topological impact of mistakes in the final prediction. We propose a new loss term that is aware of the higher-order topological features of linear structures. We also exploit a refinement pipeline that iteratively applies the same model over the previous delineation to refine the predictions at each step, while keeping the number of parameters and the complexity of the model constant. When combined with the standard pixel-wise loss, both our new loss term and an iterative refinement boost the quality of the predicted delineations, in some cases almost doubling the accuracy as compared to the same classifier trained with the binary cross-entropy alone. We show that our approach outperforms state-of-the-art methods on a wide range of data, from microscopy to aerial images.

12 November 2020 in Nature Communications
Verified
Journal Article
Open access
PDF file
NDNicola Dietler
Nicola Dietler
MMMatthias Minder
Matthias Minder
VGVojislav Gligorovski
Vojislav Gligorovski
AEAugoustina Maria Economou
Augoustina Maria Economou
DJDenis Alain Henri Lucien Joly
Denis Alain Henri Lucien Joly
SRSahand Jamal Rahi
Sahand Jamal Rahi
The identification of cell borders (‘segmentation’) in microscopy images constitutes a bottleneck for large-scale experiments. For the model organismSaccharomyces cerevisiae, current segmentation methods face challenges when cells bud, crowd, or exhibit irregular features. We present a convolutional neural network (CNN) named YeaZ, the underlying training set of high-quality segmented yeast images (>10 000 cells) including mutants, stressed cells, and time courses, as well as a graphical user interface and a web application (www.quantsysbio.com/data-and-software) to efficiently employ, test, and expand the system. A key feature is a cell-cell boundary test which avoids the need for fluorescent markers. Our CNN is highly accurate, including for buds, and outperforms existing methods on benchmark images, indicating it transfers well to other conditions. To demonstrate how efficient large-scale image processing uncovers new biology, we analyze the geometries of ≈2200 wild-type and cyclin mutant cells and find that morphogenesis control occurs unexpectedly early and gradually.

20 June 2022 in Automation in Construction
Verified
Journal Article
Open access
BRBryan German Pantoja Rosero
Bryan German Pantoja Rosero
RAR. Achanta
R. Achanta
MKM. Kozinski
M. Kozinski
PFP. Fua
P. Fua
FPF. Perez-Cruz
F. Perez-Cruz
KBKatrin Beyer
Katrin Beyer

17 March 2021 in AGU Advances
Verified
Journal Article
Open access
PDF file
NDNathan C. Dadap
Nathan C. Dadap
AHAlison M. Hoyt
Alison M. Hoyt
ACAlexander R Cobb
Alexander R Cobb
DODoruk Oner
Doruk Oner
MKMateusz Kozinski
Mateusz Kozinski
AKAlexandra G. Konings
Alexandra G. Konings
Drainage canals associated with logging and agriculture dry out organic soils in tropical peatlands, thereby threatening the viability of long‐term carbon stores due to increased emissions from decomposition, fire, and fluvial transport. In Southeast Asian peatlands, which have experienced decades of land use change, the exact extent and spatial distribution of drainage canals are unknown. This has prevented regional‐scale investigation of the relationships between drainage, land use, and carbon emissions. Here, we create the first regional map of drainage canals using high resolution satellite imagery and a convolutional neural network. We find that drainage is widespread—occurring in at least 65% of peatlands and across all land use types. Although previous estimates of peatland carbon emissions have relied on land use as a proxy for drainage, our maps show substantial variation in drainage density within land use types. Subsidence rates are 3.2 times larger in intensively drained areas than in non‐drained areas, highlighting the central role of drainage in mediating peat subsidence. Accounting for drainage canals was found to improve a subsidence prediction model by 30%, suggesting that canals contain information about subsidence not captured by land use alone. Thus, our data set can be used to improve subsidence and associated carbon emissions predictions in peatlands, and to target areas for hydrologic restoration.

6 June 2019 in IEEE Transactions on Pattern Analysis and Machine Intelligence
Verified
Journal Article
Closed access
AMAgata Mosinska
Agata Mosinska
MKMateusz Kozinski
Mateusz Kozinski
PFPascal Fua
Pascal Fua
Detection of curvilinear structures in images has long been of interest. One of the most challenging aspects of this problem is inferring the graph representation of the curvilinear network. Most existing delineation approaches first perform binary segmentation of the image and then refine it using either a set of hand-designed heuristics or a separate classifier that assigns likelihood to paths extracted from the pixel-wise prediction. In our work, we bridge the gap between segmentation and path classification by training a deep network that performs those two tasks simultaneously. We show that this approach is beneficial because it enforces consistency across the whole processing pipeline. We apply our approach on roads and neurons datasets.

1 June 2015
Verified
Conference Paper
Closed access
MKMateusz Kozinski
Mateusz Kozinski
RGRaghudeep Gadde
Raghudeep Gadde
SZSergey Zagoruyko
Sergey Zagoruyko
GOGuillaume Obozinski
Guillaume Obozinski
RMRenaud Marlet
Renaud Marlet
We present a new shape prior formalism for the segmentation of rectified facade images. It combines the simplicity of split grammars with unprecedented expressive power: the capability of encoding simultaneous alignment in two dimensions, facade occlusions and irregular boundaries between facade elements. We formulate the task of finding the most likely image segmentation conforming to a prior of the proposed form as a MAP-MRF problem over a 4-connected pixel grid, and propose an efficient optimization algorithm for solving it. Our method simultaneously segments the visible and occluding objects, and recovers the structure of the occluded facade. We demonstrate state-of-the-art results on a number of facade segmentation datasets.