Nicolás Guil
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Nicolás Guil has a total of 124 co-authors. Here are the top co-authors by number of shared publications:
- José María González-Linares
University of Malaga
32 shared publications - Julian Ramos Cozar
University of Malaga
20 shared publications - Francisco Manuel Castro
University of Malaga
19 shared publications - Manuel Jesús Marín-Jiménez
Departamento de Informática y Análisis Numérico, Edificio Einstein, Campus de Rabanales, Universidad de Córdoba, 14071 Córdoba, Spain
15 shared publications - Juan Gómez-Luna
NVIDIA, Zürich, Switzerland
14 shared publications E.L. Zapata
10 shared publications
Latest Publications
- 1 May 2026
- Published by Springer Nature
- p. 1-17
End-to-End Multitask CNN-Based Model for Palm Vein Biometrics
- 1 May 2026
- Published by Springer Nature
- p. 1-17
End-to-End Multitask CNN-Based Model for Palm Vein Biometrics
- Open Access
- 17 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Biometrics, Behavior, and Identity Science
- Vol. PP (5) , 1
HiPo-Gait: Hierarchical Pose model decoupling for Gait Recognition
Gait recognition has traditionally relied on silhouetteor appearance-based representations to analyze walking patterns. While silhouettes capture rich movement information, they are highly dependent on body shape and contours, which are potentially irrelevant to gait analysis. The human pose emerges as a more robust and semantically meaningful alternative. However, pose-based models have typically underperformed compared to silhouette-based approaches. To enhance gait feature extraction from pose representations, this paper proposes HiPo-Gait, a novel hierarchical architecture that analyzes isolated limbs of the body pose and groups them at different levels using a new local attention-based fusion module (LAF) to produce a robust hierarchical descriptor of the walking pattern. Moreover, instead of using point coordinates as input, we use heatmaps, a richer representation of the body pose. Our experimental results indicate that a) the hierarchical decoupling extracts richer features from every individual limb and b) the fusion approach optimally aggregates the limbs, outperforming the classical fusion operations. Finally, our approach achieves a percentage point increase of 3.3% over the top pose-based state-of-the-art model in CASIA-B, 21.9%, and 25.6%, respectively, in CCPG and SUSTech1K, and 1.7% improvement over the top model in the FVG-B dataset. Source code is available at https://github.com/Nico-Cubero/HiPo-Gait.
- 17 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Biometrics, Behavior, and Identity Science
- Vol. PP (5) , 1
Open AccessHiPo-Gait: Hierarchical Pose model decoupling for Gait Recognition
Gait recognition has traditionally relied on silhouetteor appearance-based representations to analyze walking patterns. While silhouettes capture rich movement information, they are highly dependent on body shape and contours, which are potentially irrelevant to gait analysis. The human pose emerges as a more robust and semantically meaningful alternative. However, pose-based models have typically underperformed compared to silhouette-based approaches. To enhance gait feature extraction from pose representations, this paper proposes HiPo-Gait, a novel hierarchical architecture that analyzes isolated limbs of the body pose and groups them at different levels using a new local attention-based fusion module (LAF) to produce a robust hierarchical descriptor of the walking pattern. Moreover, instead of using point coordinates as input, we use heatmaps, a richer representation of the body pose. Our experimental results indicate that a) the hierarchical decoupling extracts richer features from every individual limb and b) the fusion approach optimally aggregates the limbs, outperforming the classical fusion operations. Finally, our approach achieves a percentage point increase of 3.3% over the top pose-based state-of-the-art model in CASIA-B, 21.9%, and 25.6%, respectively, in CCPG and SUSTech1K, and 1.7% improvement over the top model in the FVG-B dataset. Source code is available at https://github.com/Nico-Cubero/HiPo-Gait.
- 30 July 2025
- Published by Springer Nature
- p. 214-226
Multi-scale Temporal Pose Analysis for Gait Recognition
- 30 July 2025
- Published by Springer Nature
- p. 214-226
Multi-scale Temporal Pose Analysis for Gait Recognition
- 26 May 2025
- Published by Springer Nature in International Journal of Computer Vision
- p. 1-16
Lightweight Structure-Aware Attention for Visual Understanding
- 26 May 2025
- Published by Springer Nature in International Journal of Computer Vision
- p. 1-16
Lightweight Structure-Aware Attention for Visual Understanding
- Open Access
- 15 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 87100-87111
OUMVLP-OF: Multi-View Large Population Gait Database With Dense Optical Flow and Its Performance Evaluation
This paper introduces OUMVLP-OF, a large-scale dataset that provides dense optical flow data derived from the world’s largest multi-view gait database. Unlike conventional gait datasets that primarily rely on binary silhouettes, OUMVLP-OF offers a motion-focused representation that captures fine-grained gait dynamics, enabling a richer and more detailed analysis of human movement. This way, OUMVLP-OF bridges the gap between silhouette-based and motion-based approaches, facilitating the development of novel gait recognition models that leverage both spatial and temporal information. The dataset comprises 10,307 subjects (5,114 males and 5,193 females), with gait sequences recorded from 14 distinct view angles, spanning from 0◦ to 270° in 15° intervals.A key novelty of OUMVLP-OF lies in its dual-version design: OF-V1, which preserves raw background information to encourage models to focus on subtle gait details while benefiting from natural variability as implicit data augmentation, and OF-V2, which features noise-free optical flow maps, isolating gait dynamics for precise motion analysis. This enables researchers to systematically study the impact of background noise on optical flow-based gait recognition and design more robust recognition models. By introducing the first large-scale optical flow dataset for gait analysis with both noisy and clean versions, OUMVLP-OF provides a unique benchmark for motion-based gait recognition. It advances the state of the art by enabling more robust cross-view recognition, better occlusion handling, and improved generalization across varied environments. To promote further research, OUMVLP-OF will be made publicly available at http://www.am.sanken.osaka-u.ac.jp/BiometricDB/GaitLPOF.html.
- 15 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 87100-87111
Open AccessOUMVLP-OF: Multi-View Large Population Gait Database With Dense Optical Flow and Its Performance Evaluation
This paper introduces OUMVLP-OF, a large-scale dataset that provides dense optical flow data derived from the world’s largest multi-view gait database. Unlike conventional gait datasets that primarily rely on binary silhouettes, OUMVLP-OF offers a motion-focused representation that captures fine-grained gait dynamics, enabling a richer and more detailed analysis of human movement. This way, OUMVLP-OF bridges the gap between silhouette-based and motion-based approaches, facilitating the development of novel gait recognition models that leverage both spatial and temporal information. The dataset comprises 10,307 subjects (5,114 males and 5,193 females), with gait sequences recorded from 14 distinct view angles, spanning from 0◦ to 270° in 15° intervals.A key novelty of OUMVLP-OF lies in its dual-version design: OF-V1, which preserves raw background information to encourage models to focus on subtle gait details while benefiting from natural variability as implicit data augmentation, and OF-V2, which features noise-free optical flow maps, isolating gait dynamics for precise motion analysis. This enables researchers to systematically study the impact of background noise on optical flow-based gait recognition and design more robust recognition models. By introducing the first large-scale optical flow dataset for gait analysis with both noisy and clean versions, OUMVLP-OF provides a unique benchmark for motion-based gait recognition. It advances the state of the art by enabling more robust cross-view recognition, better occlusion handling, and improved generalization across varied environments. To promote further research, OUMVLP-OF will be made publicly available at http://www.am.sanken.osaka-u.ac.jp/BiometricDB/GaitLPOF.html.
- Open Access
- 22 April 2025
- Published by Elsevier in Computer Methods and Programs in Biomedicine
- Vol. 267, 108780
FADE: Forecasting for anomaly detection on ECG
- 22 April 2025
- Published by Elsevier in Computer Methods and Programs in Biomedicine
- Vol. 267, 108780
Open AccessFADE: Forecasting for anomaly detection on ECG
- Open Access
- 28 February 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 275
Empirical study of human pose representations for gait recognition
- 28 February 2025
- Published by Elsevier in Expert Systems with Applications
- Vol. 275
Open AccessEmpirical study of human pose representations for gait recognition
- Open Access
FADE: Forecasting for Anomaly Detection on ECG
Cardiovascular diseases, a leading cause of noncommunicable disease-related deaths, require early and accurate detection to improve patient outcomes. Taking advantage of advances in machine learning and deep learning, multiple approaches have been proposed in the literature to address the challenge of detecting ECG anomalies. Typically, these methods are based on the manual interpretation of ECG signals, which is time consuming and depends on the expertise of healthcare professionals. The objective of this work is to propose a deep learning system, FADE, designed for normal ECG forecasting and anomaly detection, which reduces the need for extensive labeled datasets and manual interpretation. FADE has been trained in a self-supervised manner with a novel morphological inspired loss function. Unlike conventional models that learn from labeled anomalous ECG waveforms, our approach predicts the future of normal ECG signals, thus avoiding the need for extensive labeled datasets. Using a novel distance function to compare forecasted ECG signals with actual sensor data, our method effectively identifies cardiac anomalies. Additionally, this approach can be adapted to new contexts through domain adaptation techniques. To evaluate our proposal, we performed a set of experiments using two publicly available datasets: MIT-BIH NSR and MIT-BIH Arrythmia. The results demonstrate that our system achieves an average accuracy of 83.84% in anomaly detection, while correctly classifying normal ECG signals with an accuracy of 85.46%. Our proposed approach exhibited superior performance in the early detection of cardiac anomalies in ECG signals, surpassing previous methods that predominantly identify a limited range of anomalies. FADE effectively detects both abnormal heartbeats and arrhythmias, offering significant advantages in healthcare through cost reduction or processing of large-scale ECG data.
Open AccessFADE: Forecasting for Anomaly Detection on ECG
Cardiovascular diseases, a leading cause of noncommunicable disease-related deaths, require early and accurate detection to improve patient outcomes. Taking advantage of advances in machine learning and deep learning, multiple approaches have been proposed in the literature to address the challenge of detecting ECG anomalies. Typically, these methods are based on the manual interpretation of ECG signals, which is time consuming and depends on the expertise of healthcare professionals. The objective of this work is to propose a deep learning system, FADE, designed for normal ECG forecasting and anomaly detection, which reduces the need for extensive labeled datasets and manual interpretation. FADE has been trained in a self-supervised manner with a novel morphological inspired loss function. Unlike conventional models that learn from labeled anomalous ECG waveforms, our approach predicts the future of normal ECG signals, thus avoiding the need for extensive labeled datasets. Using a novel distance function to compare forecasted ECG signals with actual sensor data, our method effectively identifies cardiac anomalies. Additionally, this approach can be adapted to new contexts through domain adaptation techniques. To evaluate our proposal, we performed a set of experiments using two publicly available datasets: MIT-BIH NSR and MIT-BIH Arrythmia. The results demonstrate that our system achieves an average accuracy of 83.84% in anomaly detection, while correctly classifying normal ECG signals with an accuracy of 85.46%. Our proposed approach exhibited superior performance in the early detection of cardiac anomalies in ECG signals, surpassing previous methods that predominantly identify a limited range of anomalies. FADE effectively detects both abnormal heartbeats and arrhythmias, offering significant advantages in healthcare through cost reduction or processing of large-scale ECG data.
- Open Access
- 6 February 2025
- Published by Elsevier in Future Generation Computer Systems
- Vol. 167
Real-time unsupervised video object detection on the edge
- 6 February 2025
- Published by Elsevier in Future Generation Computer Systems
- Vol. 167
Open AccessReal-time unsupervised video object detection on the edge
- 25 June 2023
- Published by Springer Nature
- p. 466-478
Real-Time Unsupervised Object Localization on the Edge for Airport Video Surveillance
- 25 June 2023
- Published by Springer Nature
- p. 466-478
Real-Time Unsupervised Object Localization on the Edge for Airport Video Surveillance