Janez Krizaj
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Janez Krizaj has a total of 54 co-authors. Here are the top co-authors by number of shared publications:
- Vitomir Štruc
Faculty of Electrical Engineering, University of Ljubljana, Ljubljana, Slovenia
7 shared publications - Simon Dobrišek
Faculty of Electrical Engineering, University of Ljubljana, Ljubljana, Slovenia
6 shared publications Vitomir Struc
Faculty of Electrical Engineering, University of Ljubljana, Slovenia
6 shared publicationsPeter Peer
Faculty of Computer and Information Science, University of Ljubljana, Ljubljana, Slovenia
4 shared publicationsVitomir Štruc
Faculty of Electrical Engineering, University of Ljubljana, Ljubljana, Slovenia
4 shared publicationsPatrick J. Flynn
3 shared publications
Latest Publications
- Open Access
- 6 March 2026
- Published by University of Maribor
- p. 25-36
Benchmarking Road Network Extraction Methods for Power Distribution Spatial Planning
Accurate road network extraction from high-resolution overhead imagery is a prerequisite for corridor-aware routing and cost modelling in power-system infrastructure planning, where medium- and low-voltage lines often follow transport rights-of-way. Yet roads are thin, occluded, and cluttered, and downstream optimisation needs routable graphs, not just masks. We benchmark six representative pipelines (SAM-Road, D-LinkNet, CRESI, CU-dGCN, Sat2Graph, and U-Net-ResNet18) on SpaceNet imagery. The comparison reveals trade-offs between segmentation-centric and graph-aware designs and highlights where topology fails most often. Finally, practical guidance is provided for selecting road extractors that deliver reliable, optimisation-ready networks for power distribution corridor planning and the routing of medium- and low-voltage lines.
- 6 March 2026
- Published by University of Maribor
- p. 25-36
Open AccessBenchmarking Road Network Extraction Methods for Power Distribution Spatial Planning
Accurate road network extraction from high-resolution overhead imagery is a prerequisite for corridor-aware routing and cost modelling in power-system infrastructure planning, where medium- and low-voltage lines often follow transport rights-of-way. Yet roads are thin, occluded, and cluttered, and downstream optimisation needs routable graphs, not just masks. We benchmark six representative pipelines (SAM-Road, D-LinkNet, CRESI, CU-dGCN, Sat2Graph, and U-Net-ResNet18) on SpaceNet imagery. The comparison reveals trade-offs between segmentation-centric and graph-aware designs and highlights where topology fails most often. Finally, practical guidance is provided for selecting road extractors that deliver reliable, optimisation-ready networks for power distribution corridor planning and the routing of medium- and low-voltage lines.
- Open Access
The Neglected Group: Cognitive Discourse Markers as Signposts of Prosodic Unit Boundaries
The present paper examines and compares the role of cognitive discourse markers (DMs), such as uhm, like, or I mean, and a set of prosodic parameters as indicators of prosodic boundaries. Cognitive DMs traditionally are not studied as a separate DM group on par with the ideational, sequential, rhetorical, or interpersonal group. However, as they reflect the speaker’s mental processes during speech production, they offer an exceptional glimpse into how speakers construct their verbalisations. Along with the analysis of DMs, prosodic parameters, including pitch and intensity reset, speech rate change, and pauses, were automatically annotated to determine how well they overlapped with the manually annotated prosodic boundaries. To accommodate for the natural variability in speech, the parameters were evaluated using relative comparison methods. Among the prosodic parameters, pauses were found to overlap most often with the manually annotated prosodic boundaries. Cognitive DMs in the function of realising new information, restructuring, and emphasis indeed proved as relevant boundary indicators, however, the group of cognitive DMs as a whole fell behind the group of sequential and rhetorical DMs, which overlapped most frequently with the manually annotated prosodic boundaries.
Open AccessThe Neglected Group: Cognitive Discourse Markers as Signposts of Prosodic Unit Boundaries
The present paper examines and compares the role of cognitive discourse markers (DMs), such as uhm, like, or I mean, and a set of prosodic parameters as indicators of prosodic boundaries. Cognitive DMs traditionally are not studied as a separate DM group on par with the ideational, sequential, rhetorical, or interpersonal group. However, as they reflect the speaker’s mental processes during speech production, they offer an exceptional glimpse into how speakers construct their verbalisations. Along with the analysis of DMs, prosodic parameters, including pitch and intensity reset, speech rate change, and pauses, were automatically annotated to determine how well they overlapped with the manually annotated prosodic boundaries. To accommodate for the natural variability in speech, the parameters were evaluated using relative comparison methods. Among the prosodic parameters, pauses were found to overlap most often with the manually annotated prosodic boundaries. Cognitive DMs in the function of realising new information, restructuring, and emphasis indeed proved as relevant boundary indicators, however, the group of cognitive DMs as a whole fell behind the group of sequential and rhetorical DMs, which overlapped most frequently with the manually annotated prosodic boundaries.
- Open Access
- 31 January 2024
- Published by Elsevier in Engineering Applications of Artificial Intelligence
- Vol. 132
Deep Face Decoder: Towards understanding the embedding space of convolutional networks through visual reconstruction of deep face templates
- 31 January 2024
- Published by Elsevier in Engineering Applications of Artificial Intelligence
- Vol. 132
Open AccessDeep Face Decoder: Towards understanding the embedding space of convolutional networks through visual reconstruction of deep face templates
- 29 August 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1165-1169
Validation of Speech Data for Training Automatic Speech Recognition Systems
Recent automatic speech recognition systems are largely based on deep neural networks that need large amounts of labelled speech data to train. This can be a problem, especially for languages for which large speech databases are not available. To facilitate the construction of a speech database suitable for training automatic speech recognizers, we propose a tool that enables the validation of audio recordings from collected speech recordings. The developed tool allows the user to check the com-pliance with the predefined requirements regarding the correct audio format, the appropriate speech volume, the compatibility of the spoken text with the reference text and the suitability of the length of the non-spoken segments. The applicability of the developed tool is demonstrated by the creation of the Slovene speech corpus from audio recordings collected within the project Development of Slovene in a Digital Environment, although the tool is also suitable for all other languages supported by the used automatic speech recognizer.
- 29 August 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1165-1169
Validation of Speech Data for Training Automatic Speech Recognition Systems
Recent automatic speech recognition systems are largely based on deep neural networks that need large amounts of labelled speech data to train. This can be a problem, especially for languages for which large speech databases are not available. To facilitate the construction of a speech database suitable for training automatic speech recognizers, we propose a tool that enables the validation of audio recordings from collected speech recordings. The developed tool allows the user to check the com-pliance with the predefined requirements regarding the correct audio format, the appropriate speech volume, the compatibility of the spoken text with the reference text and the suitability of the length of the non-spoken segments. The applicability of the developed tool is demonstrated by the creation of the Slovene speech corpus from audio recordings collected within the project Development of Slovene in a Digital Environment, although the tool is also suitable for all other languages supported by the used automatic speech recognizer.
- Open Access
Making the Most of Single Sensor Information: A Novel Fusion Approach for 3D Face Recognition Using Region Covariance Descriptors and Gaussian Mixture Models
Most commercially successful face recognition systems combine information from multiple sensors (2D and 3D, visible light and infrared, etc.) to achieve reliable recognition in various environments. When only a single sensor is available, the robustness as well as efficacy of the recognition process suffer. In this paper, we focus on face recognition using images captured by a single 3D sensor and propose a method based on the use of region covariance matrixes and Gaussian mixture models (GMMs). All steps of the proposed framework are automated, and no metadata, such as pre-annotated eye, nose, or mouth positions is required, while only a very simple clustering-based face detection is performed. The framework computes a set of region covariance descriptors from local regions of different face image representations and then uses the unscented transform to derive low-dimensional feature vectors, which are finally modeled by GMMs. In the last step, a support vector machine classification scheme is used to make a decision about the identity of the input 3D facial image. The proposed framework has several desirable characteristics, such as an inherent mechanism for data fusion/integration (through the region covariance matrixes), the ability to explore facial images at different levels of locality, and the ability to integrate a domain-specific prior knowledge into the modeling procedure. Several normalization techniques are incorporated into the proposed framework to further improve performance. Extensive experiments are performed on three prominent databases (FRGC v2, CASIA, and UMB-DB) yielding competitive results.
Open AccessMaking the Most of Single Sensor Information: A Novel Fusion Approach for 3D Face Recognition Using Region Covariance Descriptors and Gaussian Mixture Models
Most commercially successful face recognition systems combine information from multiple sensors (2D and 3D, visible light and infrared, etc.) to achieve reliable recognition in various environments. When only a single sensor is available, the robustness as well as efficacy of the recognition process suffer. In this paper, we focus on face recognition using images captured by a single 3D sensor and propose a method based on the use of region covariance matrixes and Gaussian mixture models (GMMs). All steps of the proposed framework are automated, and no metadata, such as pre-annotated eye, nose, or mouth positions is required, while only a very simple clustering-based face detection is performed. The framework computes a set of region covariance descriptors from local regions of different face image representations and then uses the unscented transform to derive low-dimensional feature vectors, which are finally modeled by GMMs. In the last step, a support vector machine classification scheme is used to make a decision about the identity of the input 3D facial image. The proposed framework has several desirable characteristics, such as an inherent mechanism for data fusion/integration (through the region covariance matrixes), the ability to explore facial images at different levels of locality, and the ability to integrate a domain-specific prior knowledge into the modeling procedure. Several normalization techniques are incorporated into the proposed framework to further improve performance. Extensive experiments are performed on three prominent databases (FRGC v2, CASIA, and UMB-DB) yielding competitive results.
- Open Access
- 23 October 2019
- Published by Springer Nature in Neural Computing and Applications
- Vol. 32 (24) , 17909-17926
Simultaneous multi-descent regression and feature learning for facial landmarking in depth images
Face alignment (or facial landmarking) is an important task in many face-related applications, ranging from registration, tracking, and animation to higher-level classification problems such as face, expression, or attribute recognition. While several solutions have been presented in the literature for this task so far, reliably locating salient facial features across a wide range of posses still remains challenging. To address this issue, we propose in this paper a novel method for automatic facial landmark localization in 3D face data designed specifically to address appearance variability caused by significant pose variations. Our method builds on recent cascaded regression-based methods to facial landmarking and uses a gating mechanism to incorporate multiple linear cascaded regression models each trained for a limited range of poses into a single powerful landmarking model capable of processing arbitrary-posed input data. We develop two distinct approaches around the proposed gating mechanism: (1) the first uses a gated multiple ridge descent mechanism in conjunction with established (hand-crafted) histogram of gradients features for face alignment and achieves state-of-the-art landmarking performance across a wide range of facial poses and (2) the second simultaneously learns multiple-descent directions as well as binary features that are optimal for the alignment tasks and in addition to competitive landmarking results also ensures extremely rapid processing. We evaluate both approaches in rigorous experiments on several popular datasets of 3D face images, i.e., the FRGCv2 and Bosphorus 3D face datasets and image collections F and G from the University of Notre Dame. The results of our evaluation show that both approaches compare favorably to the state-of-the-art, while exhibiting considerable robustness to pose variations.
- 23 October 2019
- Published by Springer Nature in Neural Computing and Applications
- Vol. 32 (24) , 17909-17926
Open AccessSimultaneous multi-descent regression and feature learning for facial landmarking in depth images
Face alignment (or facial landmarking) is an important task in many face-related applications, ranging from registration, tracking, and animation to higher-level classification problems such as face, expression, or attribute recognition. While several solutions have been presented in the literature for this task so far, reliably locating salient facial features across a wide range of posses still remains challenging. To address this issue, we propose in this paper a novel method for automatic facial landmark localization in 3D face data designed specifically to address appearance variability caused by significant pose variations. Our method builds on recent cascaded regression-based methods to facial landmarking and uses a gating mechanism to incorporate multiple linear cascaded regression models each trained for a limited range of poses into a single powerful landmarking model capable of processing arbitrary-posed input data. We develop two distinct approaches around the proposed gating mechanism: (1) the first uses a gated multiple ridge descent mechanism in conjunction with established (hand-crafted) histogram of gradients features for face alignment and achieves state-of-the-art landmarking performance across a wide range of facial poses and (2) the second simultaneously learns multiple-descent directions as well as binary features that are optimal for the alignment tasks and in addition to competitive landmarking results also ensures extremely rapid processing. We evaluate both approaches in rigorous experiments on several popular datasets of 3D face images, i.e., the FRGCv2 and Bosphorus 3D face datasets and image collections F and G from the University of Notre Dame. The results of our evaluation show that both approaches compare favorably to the state-of-the-art, while exhibiting considerable robustness to pose variations.
- Open Access
Simultaneous regression and feature learning for facial landmarking
Face alignment (or facial landmarking) is an important task in many face-related applications, ranging from registration, tracking and animation to higher-level classification problems such as face, expression or attribute recognition. While several solutions have been presented in the literature for this task so far, reliably locating salient facial features across a wide range of posses still remains challenging. To address this issue, we propose in this paper a novel method for automatic facial landmark localization in 3D face data designed specifically to address appearance variability caused by significant pose variations. Our method builds on recent cascaded-regression-based methods to facial landmarking and uses a gating mechanism to incorporate multiple linear cascaded regression models each trained for a limited range of poses into a single powerful landmarking model capable of processing arbitrary posed input data. We develop two distinct approaches around the proposed gating mechanism: i) the first uses a gated multiple ridge descent (GRID) mechanism in conjunction with established (hand-crafted) HOG features for face alignment and achieves state-of-the-art landmarking performance across a wide range of facial poses, ii) the second simultaneously learns multiple-descent directions as well as binary features (SMUF) that are optimal for the alignment tasks and in addition to competitive landmarking results also ensures extremely rapid processing. We evaluate both approaches in rigorous experiments on several popular datasets of 3D face images, i.e., the FRGCv2 and Bosphorus 3D Face datasets and image collections F and G from the University of Notre Dame. The results of our evaluation show that both approaches are competitive in comparison to the state-of-the-art, while exhibiting considerable robustness to pose variations.
Open AccessSimultaneous regression and feature learning for facial landmarking
Face alignment (or facial landmarking) is an important task in many face-related applications, ranging from registration, tracking and animation to higher-level classification problems such as face, expression or attribute recognition. While several solutions have been presented in the literature for this task so far, reliably locating salient facial features across a wide range of posses still remains challenging. To address this issue, we propose in this paper a novel method for automatic facial landmark localization in 3D face data designed specifically to address appearance variability caused by significant pose variations. Our method builds on recent cascaded-regression-based methods to facial landmarking and uses a gating mechanism to incorporate multiple linear cascaded regression models each trained for a limited range of poses into a single powerful landmarking model capable of processing arbitrary posed input data. We develop two distinct approaches around the proposed gating mechanism: i) the first uses a gated multiple ridge descent (GRID) mechanism in conjunction with established (hand-crafted) HOG features for face alignment and achieves state-of-the-art landmarking performance across a wide range of facial poses, ii) the second simultaneously learns multiple-descent directions as well as binary features (SMUF) that are optimal for the alignment tasks and in addition to competitive landmarking results also ensures extremely rapid processing. We evaluate both approaches in rigorous experiments on several popular datasets of 3D face images, i.e., the FRGCv2 and Bosphorus 3D Face datasets and image collections F and G from the University of Notre Dame. The results of our evaluation show that both approaches are competitive in comparison to the state-of-the-art, while exhibiting considerable robustness to pose variations.
- 1 July 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-8
Localization of Facial Landmarks in Depth Images Using Gated Multiple Ridge Descent
A novel method for automatic facial landmark localization is presented. The method builds on the supervised descent framework, which was shown to successfully localize landmarks in the presence of large expression variations and mild occlusions, but struggles when localizing landmarks on faces with large pose variations. We propose an extension of the supervised descent framework that trains multiple descent maps and results in increased robustness to pose variations. The performance of the proposed method is demonstrated on the Bosphorus, the FRGC and the UND data sets for the problem of facial landmark localization from 3D data. Our experimental results show that the proposed method exhibits increased robustness to pose variations, while retaining high performance in the case of expression and occlusion variations.
- 1 July 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-8
Localization of Facial Landmarks in Depth Images Using Gated Multiple Ridge Descent
A novel method for automatic facial landmark localization is presented. The method builds on the supervised descent framework, which was shown to successfully localize landmarks in the presence of large expression variations and mild occlusions, but struggles when localizing landmarks on faces with large pose variations. We propose an extension of the supervised descent framework that trains multiple descent maps and results in increased robustness to pose variations. The performance of the proposed method is demonstrated on the Bosphorus, the FRGC and the UND data sets for the problem of facial landmark localization from 3D data. Our experimental results show that the proposed method exhibits increased robustness to pose variations, while retaining high performance in the case of expression and occlusion variations.
- Open Access
To Frontalize or Not To Frontalize: Do We Really Need Elaborate Pre-processing To Improve Face Recognition?
Face recognition performance has improved remarkably in the last decade. Much of this success can be attributed to the development of deep learning techniques such as convolutional neural networks (CNNs). While CNNs have pushed the state-of-the-art forward, their training process requires a large amount of clean and correctly labelled training data. If a CNN is intended to tolerate facial pose, then we face an important question: should this training data be diverse in its pose distribution, or should face images be normalized to a single pose in a pre-processing step? To address this question, we evaluate a number of popular facial landmarking and pose correction algorithms to understand their effect on facial recognition performance. Additionally, we introduce a new, automatic, single-image frontalization scheme that exceeds the performance of current algorithms. CNNs trained using sets of different pre-processing methods are used to extract features from the Point and Shoot Challenge (PaSC) and CMU Multi-PIE datasets. We assert that the subsequent verification and recognition performance serves to quantify the effectiveness of each pose correction scheme.
Open AccessTo Frontalize or Not To Frontalize: Do We Really Need Elaborate Pre-processing To Improve Face Recognition?
Face recognition performance has improved remarkably in the last decade. Much of this success can be attributed to the development of deep learning techniques such as convolutional neural networks (CNNs). While CNNs have pushed the state-of-the-art forward, their training process requires a large amount of clean and correctly labelled training data. If a CNN is intended to tolerate facial pose, then we face an important question: should this training data be diverse in its pose distribution, or should face images be normalized to a single pose in a pre-processing step? To address this question, we evaluate a number of popular facial landmarking and pose correction algorithms to understand their effect on facial recognition performance. Additionally, we introduce a new, automatic, single-image frontalization scheme that exceeds the performance of current algorithms. CNNs trained using sets of different pre-processing methods are used to extract features from the Point and Shoot Challenge (PaSC) and CMU Multi-PIE datasets. We assert that the subsequent verification and recognition performance serves to quantify the effectiveness of each pose correction scheme.