Publications by Mateusz Kozinski

Generating LOD3 building models from structure-from-motion and semantic segmentation

Pascal Fua, Radhakrishna Achanta, Katrin Beyer, Mateusz Kozinski, Fernando Perez Cruz, Bryan German Pantoja Rosero

This paper describes a pipeline for automatically generating level of detail (LOD) models (digital twins), specifically LOD2 and LOD3, from free-standing buildings. Our approach combines structure from motion (SfM) with deep-learning-based segmentation tec ...

2022

Adjusting the Ground Truth Annotations for Connectivity-Based Learning to Delineate

Pascal Fua, Leonardo Citraro, Mateusz Kozinski, Doruk Oner

Deep learning-based approaches to delineating 3D structure depend on accurate annotations to train the networks. Yet in practice, people, no matter how conscientious, have trouble precisely delineating in 3D and on a large scale, in part because the data i ...

Institute of Electrical and Electronics Engineers2022

Persistent Homology with Improved Locality Information for more Effective Delineation

Kathryn Hess Bellwald, Pascal Fua, Mateusz Kozinski, Adélie Eliane Garin, Doruk Oner

Persistent Homology (PH) has been successfully used to train networks to detect curvilinear structures and to improve the topological quality of their results. However, existing methods are very global and ignore the location of topological features. In th ...

2022

Promoting Connectivity of Network-Like Structures by Enforcing Region Separation

Pascal Fua, Leonardo Citraro, Mateusz Kozinski, Doruk Oner

We propose a novel, connectivity-oriented loss function for training deep convolutional networks to reconstruct network-like structures, like roads and irrigation canals, from aerial images. The main idea behind our loss is to express the connectivity of r ...

2022

Enforcing Connectivity of 3D Linear Structures Using Their 2D Projections

Pascal Fua, Mateusz Kozinski, Doruk Oner, Hussein Fadl Hassan Hassan Osman

Many biological and medical tasks require the delineation of 3D curvilinear structures such as blood vessels and neurites from image volumes. This is typically done using neural networks trained by minimizing voxel-wise loss functions that do not capture t ...

SPRINGER INTERNATIONAL PUBLISHING AG2022

Drainage Canals in Southeast Asian Peatlands Increase Carbon Emissions

Mateusz Kozinski, Doruk Oner

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 ...

AMER GEOPHYSICAL UNION2021

Tracing in 2D to reduce the annotation effort for 3D deep delineation of linear structures

Pascal Fua, Mathieu Salzmann, Agata Justyna Mosinska, Mateusz Kozinski

The difficulty of obtaining annotations to build training databases still slows down the adoption of recent deep learning approaches for biomedical image analysis. In this paper, we show that we can train a Deep Net to perform 3D volumetric delineation giv ...

2020

A convolutional neural network segments yeast microscopy images with high accuracy

Nicola Dietler, Matthias Minder, Ahmad Sadeghi, Mateusz Kozinski, Sahand Jamal Rahi, Vojislav Gligorovski, Anne-Florence Raphaëlle Bitbol, Martin Weigert

The identification of cell borders ('segmentation') in microscopy images constitutes a bottleneck for large-scale experiments. For the model organism Saccharomyces cerevisiae, current segmentation methods face challenges when cells bud, crowd, or exhibit i ...

Nature Research2020

Joint Segmentation and Path Classification of Curvilinear Structures

Pascal Fua, Agata Justyna Mosinska, Mateusz Kozinski

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 segment ...

2020

Towards Reliable Evaluation of Algorithms for Road Network Reconstruction from Aerial Images

Pascal Fua, Leonardo Citraro, Mateusz Kozinski

Existing connectivity-oriented performance measures rank road delineation algorithms inconsistently, which makes it difficult to decide which one is best for a given application. We show that these inconsistencies stem from design flaws that make the metri ...

2020

TopoAL: An Adversarial Learning Approach for Topology-Aware Road Segmentation

Pascal Fua, Leonardo Citraro, Mateusz Kozinski, Subeesh Vasu

Most state-of-the-art approaches to road extraction from aerial images rely on a CNN trained to label road pixels as foreground and remainder of the image as background. The CNN is usually trained by minimizing pixel-wise losses, which is less than ideal t ...

2020

Learning to Segment 3D Linear Structures Using Only 2D Annotations

Pascal Fua, Mathieu Salzmann, Agata Justyna Mosinska, Mateusz Kozinski

We propose a loss function for training a Deep Neural Network (DNN) to segment volumetric data, that accommodates ground truth annotations of 2D projections of the training volumes, instead of annotations of the 3D volumes themselves. In consequence, we si ...

SPRINGER INTERNATIONAL PUBLISHING AG2018

Beyond the Pixel-Wise Loss for Topology-Aware Delineation

Pascal Fua, Pablo Marquez Neila, Agata Justyna Mosinska, Mateusz Kozinski

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 architectur ...

2018