SciTePress - Affiliation Profile

IEEE, United States

Context (please use complete words only)

Field

Sort by

14 Papers

A Comparative Study of CNNs and Vision-Language Models for Chart Image Classification

Topics: Neural Networks; Vision and Perception; Visualization; Deep Learning

Authors: Bruno Côme ^{1

;

2} , Maxime Devanne ¹ , Jonathan Weber ¹ and Germain Forestier ¹

Affiliations: ¹ IRIMAS, University of Haute-Alsace, France , ² Duke, Saint-Paul, La Reunion, France

Digital Object Identifier: 10.5220/0013374500003890

Proceedings of ICAART 2025, ISBN: 978-989-758-737-5

Quick Abstract

Chart image classification is a critical task in automating data extraction and interpretation from visualizations, which are widely used in domains such as business, research, and education. In this paper, we evaluate the performance of Convolutional Neural Networks (CNNs) and Vision-Language Models (VLMs) for this task, given their increasing use in various image classification and comprehension tasks. We constructed a diverse dataset of 25 chart types, each containing 1,000 images, and trained multiple CNN architectures while also assessing the zero-shot generalization capabilities of pre-trained VLMs. Our results demonstrate that CNNs, when trained specifically for chart classification, outperform VLMs, which nonetheless show promising potential without the need for task-specific training. These findings underscore the importance of CNNs in chart classification while highlighting the unexplored potential of VLMs with further fine-tuning, making this task crucial for advancing automated data visualization analysis.

Deep Neural Network Architectures for Advanced Hiking Map Generation

Topics: Deep Learning and Neural Networks; Data Mining and Algorithms for Big Data

Authors: Olivier Schirm ¹ , Maxime Devanne ² , Jonathan Weber ² , Arnaud Lecus ¹ , Germain Forestier ² and Cédric Wemmert ³

Affiliations: ¹ Visorando, Soultz-Haut-Rhin, France , ² IRIMAS, University of Haute Alsace, Mulhouse, France , ³ ICube, University of Strasbourg, Strasbourg, France

Digital Object Identifier: 10.5220/0013351700003905

Proceedings of ICPRAM 2025, ISBN: 978-989-758-730-6

Quick Abstract

The automation of hiking map generation using deep learning represents a pivotal advancement in geospatial analysis. This study investigates the application of neural network architectures to derive accurate hiking maps from GPS trajectory data, exclusively collected via the Visorando mobile application. By exploring the utility of 17 distinct raster features derived from geospatial data, we identify the heatmap as the most effective input for mapping intricate trail networks, achieving superior performance across accuracy, segmentation, and connectivity metrics. Among various architectures evaluated, HRNet emerged as the most efficient model, demonstrating exceptional results when combined with optimal input features, significantly outperforming state-of-the-art approaches in intersection detection and trail segmentation. This research introduces a novel framework for converting vector-based GPS traces into rasterized data suitable for convolutional neural networks, overcoming challenges like noisy inputs and terrain variability. The findings establish a new benchmark for efficiency and accuracy in hiking map generation, offering valuable insights for the broader field of automated map inference reliant on GPS data. Future work will explore direct spatio-temporal processing of vector data, eliminating raster conversion for enhanced scalability and precision.

Enhancing Time Series Classification with Self-Supervised Learning

Topics: Neural Networks; Machine Learning; Deep Learning

Authors: Ali Ismail-Fawaz , Maxime Devanne , Jonathan Weber and Germain Forestier

Affiliation: IRIMAS, Université Haute-Alsace, Mulhouse, France

Digital Object Identifier: 10.5220/0011611300003393

Proceedings of ICAART 2023, ISBN: 978-989-758-623-1

Quick Abstract

Self-Supervised Learning (SSL) is a range of Machine Learning techniques having the objective to reduce the amount of labeled data required to train a model. In Deep Learning models, SSL is often implemented using specific loss functions relying on pretext tasks leveraging from unlabelled data. In this paper, we explore SSL for the specific task of Time Series Classification (TSC). In the last few years, dozens of Deep Learning architectures were proposed for TSC. However, they almost exclusively rely on the traditional training step involving only labeled data in sufficient numbers. In order to study the potential of SSL for TSC, we propose the TRIplet Loss In TimE (TRILITE) which relies on an existing triplet loss mechanism and which does not require labeled data. We explore two use cases. In the first one, we evaluate the interest of TRILITE to boost a supervised classifier when very few labeled data are available. In the second one, we study the use of TRILITE in the context of semi-supervised learning, when both labeled and unlabeled data are available. Experiments performed on 85 datasets from the UCR archive reveal interesting results on some datasets in both use cases.

Ensemble Clustering for Histopathological Images Segmentation using Convolutional Autoencoders

Topics: Medical Image Applications; Deep Learning for Visual Understanding

Authors: Ilias Rmouque ¹ , Maxime Devanne ² , Jonathan Weber ² , Germain Forestier ² and Cédric Wemmert ¹

Affiliations: ¹ ICube, University of Strasbourg, France , ² IRIMAS, University of Haute-Alsace, France

Digital Object Identifier: 10.5220/0010835300003124

Proceedings of VISAPP 2022/VISIGRAPP, ISBN: 978-989-758-555-5

Quick Abstract

Unsupervised deep learning using autoencoders has shown excellent results in image analysis and computer vision. However, only few studies have been presented in the field of digital pathology, where proper labelling of the objects of interest is a particularly costly and difficult task. Thus, having a first fully unsupervised segmentation could greatly help in the analysis process of such images. In this paper, many architectures of convolutional autoencoders have been compared to study the influence of three main hyperparameters: (1) number of convolutional layers, (2) number of convolutions in each layer and (3) size of the latent space. Different clustering algorithms are also compared and we propose a new way to obtain more precise results by applying ensemble clustering techniques which consists in combining multiple clustering results.

Multi-layer Fog Computing Framework for Constrained LoRa Networks Intended for Water Quality Monitoring and Precision Agriculture Systems

Topics: Internet of Things; Coverage, Connectivity and Deployment Issues; Wireless Communication; Wi-Fi, Wimax and Bluetooth; Wireless Channels

Authors: Laura García ^{1

;

2} , Jose M. Jimenez ^{1

;

2} , Sandra Sendra ¹ , Jaime Lloret ¹ and Pascal Lorenz ²

Affiliations: ¹ Instituto de Investigación para la Gestión Integrada de Zonas Costeras, Universitat Politècnica de València, C/ Paranimf nº 1, Grao de Gandía – Gandia, Valencia, Spain , ² Network and Telecommunication Research Group, University of Haute Alsace, 34 rue du Grillenbreit, 68008, Colmar, France

Digital Object Identifier: 10.5220/0010618300460055

Proceedings of WINSYS 2021, ISBN: 978-989-758-529-6

Quick Abstract

As the population of the world keeps increasing, it is necessary for the agriculture to adopt technologies that improve the production and optimize re-sources such as water. This has been done by introducing IoT devices, which has led to smart agriculture or precision agriculture. However, due to the remoteness of the fields, the communication of these devices needs to be per-formed with technologies such as LoRa that has limitations on the amount of data and the number of messages that can be forwarded. Furthermore, as there is no connection to the electric grid, optimizing the energy consumption is a necessity. In this paper, we present a multi-layer fog computing framework for a water quality monitoring and precision agriculture system. Data aggregation techniques are applied at the algorithms provided for the different layers to reduce the amount of data and the number of messages forwarded to the data center so as to improve the performance of the constrained LoRa network and reduce the energy consumption. Furthermore, the added decision-making provides fault-tolerance to the system if the connection to the Data Center is not available. Simulations were performed for different functioning modes. Results show a reduction of the 80% in the amount of transmitted data and a reduction of 85.33% in the number of for-warded messages for the most restrictive functioning mode.

Time-frequency Features for sEMG Signals Classification

Topics: Medical Signal Acquisition, Analysis and Processing; Pattern Recognition & Machine Learning for Biosignal Data; Frequency Analysis & Wavelet Transform

Authors: Somar Karheily , Ali Moukadem , Jean-Baptiste Courbot and Djaffar Ould Abdeslam

Affiliation: Institute IRIMAS, University of Haute-Alsace, Mulhouse, France

Digital Object Identifier: 10.5220/0008971902440249

Proceedings of BIOSIGNALS 2020/BIOSTEC, ISBN: 978-989-758-398-8

Quick Abstract

This paper proposes a new approach for the identification of hand movements in order to control prosthetic hand. sEMG signals were used to identify movements by using two time frequency transforms: Short Time Fourier Transform and Stockwell transform. Then, we apply Singular Value Decomposition (SVD) to decrease the features dimension and to form the final features’ vector. These extracted features were used by two kinds of classifiers: K nearest neighbours and linear discriminant analysis. Finally, we numerically study these methods on a database of 10 subjects and 17 hand gestures.

A Domain Specific Platform for Engineering Well Founded Measurement Applications

Topics: Model-Driven Engineering; Requirements Elicitation and Specification; Mobile Technologies and Applications; Quality Control and Assurance; Human- Computer Interaction and Usability Engineering

Authors: Florent Bourgeois ¹ , Philippe Studer ² , Bernard Thirion ² and Jean-Marc Perronne ²

Affiliations: ¹ Université de Haute Alsace and Actimage GmbH, France , ² Université de Haute Alsace, France

Digital Object Identifier: 10.5220/0005512103090318

Proceedings of ICSOFT-EA 2015/ICSOFT, ISBN: 978-989-758-114-4

Quick Abstract

Mobile platforms, such as smartphones, are now embedding more processing and communication capabilities than ever. They offer generally a set of standard built-in sensors to measure their surroundings and potentially increase their knowledge about the environment. Moreover their communication capabilities allow easy access to external devices and remotely accessible sensing nodes or more general services. Nevertheless, despite their obvious ability to provide rich data visualization, only a few applications propose using mobile platforms as a flexible and user-friendly measuring process assistant. This paper proposes the description of a system able to model and design mobile and well-founded domain specific measuring processes, supporting physical as well as non-physical quantities. The soundness of the application and its conformance to metrology rules is ensured through the use of quantities semantic, dimensional analysis and adherence to the representational theory of measurement. The conformance verification gives to non-metrology specialists the ability to design and configure rigorous mobile applications dedicated to assist an end-user in its usual and specific measuring needs and habits, while limiting erroneous results due to manipulation errors.

Related Ontology Subjects/Areas/Topics (13)

Biomedical Engineering	Mobile Technologies	Neurotechnology, Electronics and Informatics
Cross-Feeding between Data and Software Engineering	Mobile Technologies for Healthcare Applications	Requirements Elicitation and Specification
Distributed and Mobile Software Systems	Model-Driven Engineering	Software Engineering
Health Engineering and Technology Applications	Neural Rehabilitation	Software Engineering Methods and Techniques
Health Information Systems

Analysis of the Electromechanical Activity of the Heart from Synchronized ECG and PCG Signals of Subjects Under Stress

Topics: Medical Signal Acquisition, Analysis and Processing; Wearable Sensors and Systems; Cardiovascular Signals

Authors: Ana Castro ¹ , Ali Moukadem ² , Samuel Schmidt ³ , Alain Dieterlen ² and Miguel T. Coimbra ¹

Affiliations: ¹ Universidade do Porto, Portugal , ² University of Haute Alsace, France , ³ Aalborg University, Denmark

Digital Object Identifier: 10.5220/0005202400490056

Proceedings of BIOSIGNALS 2015/BIOSTEC, ISBN: 978-989-758-069-7

Quick Abstract

In this exploratory study we propose to analyze, in healthy adult volunteers, the heart electrical (electrocardiogram, ECG) and mechanical (phonocardiogram, PCG) activity during exercise. Heart sounds amplitude, frequency content, and RS2, may be important features in the non-invasive assessment of heart activity, such as for the estimation of cardiac output and blood pressure. Nine healthy volunteers were monitored with ECG and PCG simultaneously, under a stress test. After each workload level a 10 s window of signal was collected. PCG first (S1) and second (S2) heart sounds were manually annotated, based on time of QRS complex occurrence. A QRS detector was implemented to detect the QRS complex, and time intervals between electrical and mechanical events. Extracted features were analyzed in relation to heart rate (HR), including RS2, S1 and S2 amplitudes, and high frequency content of S1 and S2. Spearman correlation was used. Changes between baseline and maximum workload stage/HR for each volunteer were analyzed. Significant correlation was observed between HR, and all characteristics extracted (P<0.01). There was a clear difference between all variables from baseline to maximum workload level: with increasing workload/HR heart sounds amplitude increased (more pronounced in S1), RS2 decreased, and high frequency content of S2 decreased in relation to the high frequency content of S1, demonstrating that dynamic cardiovascular relations are individualized during cardiac stress and that assumptions for resting conditions may not be assumed.

Related Ontology Subjects/Areas/Topics (11)

Applications and Services	Computer Vision, Visualization and Computer Graphics	Medical Image Detection, Acquisition, Analysis and Processing
Biomedical Engineering	Devices	Physiological Computing Systems
Biomedical Signal Processing	Health Information Systems	Wearable Sensors and Systems
Cardiovascular Signals	Human-Computer Interaction

Autonomous Cardiac Diagnostic based on Synchronized ECG and PCG Signal

Authors: Z. Bouguila ¹ , A. Moukadem ¹ , A. Dieterlen ¹ , A. Ahmed Benyahia ² , A. Hajjam ³ , S. Talha ⁴ and E. Andres ⁴

Affiliations: ¹ University of Haute Alsace, France , ² Université de Technologie Belfort Montbéliard and Newel, France , ³ Université de Technologie Belfort Montbéliard, France , ⁴ Hôpital Civil, France

Proceedings of DCBIOSTEC 2014/BIOSTEC, ISBN: Not Available

Quick Abstract

Despite considerable advances in medical therapy, heart failure remains a substantial burden of mortality and economic cost. This is tangibly seen by the rising number of healthcare systems that are adopting telemedicine, by the development of industry investments in telemedicine products and involvement of government in project delivery. The main objective of the telemedicine E-care (www.projet-e-care.fr) project is to greatly contribute to the enhancement of the remote patient monitoring, expanding the possibilities for lifesaving care. In this project, oriented for heart failure patients with NYHA (class III) severity, a smart platform is adopted for home monitoring using noninvasive standard sensors and specific cardiac devices (ElectroCardioGram and PhonoCardioGram). This thesis contribution, with our partners, is to enhance the signal processing part for accurate diagnostic. Although ECG and PCG signals play important roles in heart abnormality detection, diagnosis based on this signal individually cannot detect most cases of heart symptoms, like detection of S4 pathological heart sound. Previously in our laboratory, time-frequency methods showed here potential on PCG segmentation and classification; however these methods can suffer from poor energy concentration in time frequency domain. Hence the first theoretical thesis objective is to optimize the energy concentration of Stockwell time-frequency transform. The second objective, based on the first results, is to combine by time-frequency correlation the information contained in PCG and ECG. This original approach will help us extracting features to classify accurately heart disease. All these methods should be validated by an experimented cardiologist during a measure campaign in Strasbourg University Hospital.

Fast Segmentation for Texture-based Cartography of whole Slide Images

Topics: Segmentation and Grouping; Color and Texture Analyses; Medical Image Applications

Authors: Grégory Apou ¹ , Benoît Naegel ¹ , Germain Forestier ² , Friedrich Feuerhake ³ and Cédric Wemmert ¹

Affiliations: ¹ University of Strasbourg, France , ² University of Haute Alsace, France , ³ Hannover Medical School, Germany

Digital Object Identifier: 10.5220/0004687403090319

Proceedings of VISAPP 2014/VISIGRAPP, ISBN: 978-989-758-003-1

Quick Abstract

In recent years, new optical microscopes have been developed, providing very high spatial resolution images called Whole Slide Images (WSI). The fast and accurate display of such images for visual analysis by pathologists and the conventional automated analysis remain challenging, mainly due to the image size (sometimes billions of pixels) and the need to analyze certain image features at high resolution. To propose a decision support tool to help the pathologist interpret the information contained by the WSI, we present a new approach to establish an automatic cartography of WSI in reasonable time. The method is based on an original segmentation algorithm and on a supervised multiclass classification using a textural characterization of the regions computed by the segmentation. Application to breast cancer WSI shows promising results in terms of speed and quality.

Related Ontology Subjects/Areas/Topics (6)

Applications and Services		Computer Vision, Visualization and Computer Graphics		Medical Image Applications
Color and Texture Analyses		Image and Video Analysis		Segmentation and Grouping

Context (please use complete words only)

Field

Sort by

Topic(s):

Title:
ISBN:
Year:
Acronym:
Subject:

Title:
Author:
Affiliation:
Subject:

Name:
Affiliation:
Country:
Conference:
Subject:

Name:
Country:
Subject:

Affiliation

Filter listed results, by