IFIP Open Access Digital Library
Not a member yet
22614 research outputs found
Sort by
3D Object Reconstruction with Deep Learning
No full textPart 2: Image UnderstandingInternational audienceRecent advancements and breakthroughs in deep learning have accelerated the rapid development in the field of computer vision. Having recorded a huge success in 2D object perception and detection, a lot of progress has also been made in 3D object reconstruction. Since humans can infer and relate better with 3D world images by just a single view 2D image of the object, it is necessary to train computers to think in 3D to achieve some key applications of computer vision. The use of deep learning in 3D object reconstruction of single-view images is rapidly evolving and recording significant results. In this research, we explore the Facebook well-known hybrid approach called Mesh R-CNN that combines voxel generation and triangular mesh reconstruction to generate 3D mesh structure of an object from a 2D single-view image. Although the reconstruction of objects with varying geometry and topology was achieved by Mesh R-CNN, the mesh quality was affected due to topological errors like self-intersection, causing non-smooth and rough mesh generation. In this research, Mesh R-CNN with Laplacian Smoothing (Mesh R-CNN-LS) was proposed to use the Laplacian smoothing and regularization algorithm to refine the non-smooth and rough mesh. The proposed Mesh R-CNN-LS helps to constrain the triangular deformation and generate a better and smoother 3D mesh. The proposed Mesh R-CNN-LS was compared with the original Mesh R-CNN on the Pix3D dataset and it showed better performance in terms of the loss and average precision score
Dynamic Parameter Estimation for Mixtures of Plackett-Luce Models
No full textPart 1: Pattern RecognitionInternational audienceTraditional parameter estimation algorithms rely on static datasets whose data remain constant during program execution. However, in the real-world scenario, rank data often updates in real-time, e.g., when users perform operations, such as submitting or withdrawing rankings. This dynamic nature of rank data poses challenges for applying traditional algorithms. To address this issue, we propose parameter estimation algorithms tailored for structured partial rankings based on dynamic datasets in this paper. These dynamic datasets can be classified as extended datasets and compressed datasets. To handle each dataset type, we introduce the extension preference learning algorithm and the compression preference learning algorithm based on GMM and Elsr algorithms, respectively. These algorithms ensure a relatively consistent dataset size over time, balancing accuracy and efficiency. Experimental results conducted in this paper compare the accuracy, efficiency, and stability of various algorithms using synthetic datasets, Sushi datasets, and Irish datasets, which demonstrate the effectiveness of our proposed algorithm in real-world scenarios
A Bibliometric Perspective of Integrating Labor Flexibility in Workload Control
No full textPart 2: Human-centred Manufacturing and Logistics Systems Design and Management for the Operator 5.0International audienceIn an era assessed by quick technological advancement and shifting work paradigms, the integration of labor flexibility (LF) within Workload Control (WLC) systems presents a critical yet underexplored facet of industrial operation. This research embarks on a comprehensive bibliometric and systematic literature network analysis crossing a decade of studies from 2014 to 2024, uncovering pivotal contributions and identifying prevalent themes in LF and WLC. Although our inquiry reveals an intensifying interest in this field, particularly during the COVID-19 pandemic, it discloses a noticeable research shortfall in empirical explorations of LF’s role within various order release methodologies. Addressing this gap, the study brings to light the growing importance of human dynamics, such as learning curves and worker heterogeneity, in optimizing WLC. Synthesizing the most significant scholarly works, the paper points out the urgency of adopting cross-disciplinary approaches to enrich future research attempts
Game-Based Design of a Human-Machine Collaboration Monitoring System
No full textPart 2: Human-centred Manufacturing and Logistics Systems Design and Management for the Operator 5.0International audienceIn a human-machine collaboration scenario, identifying a specific use case can be challenging due to the wide range of potential applications and interactions. Additionally, effective monitoring of the behavior of both human and machine agents during this collaboration poses significant challenges. The developed game-based process enables the analysis of the behaviours, leading to improved efficiency and collaboration. Indicators such as agent utilization and waiting times serve as valuable metrics to represent the quality of collaboration. This paper presents a setting in the Industry 5.0 laboratory, where monitoring and evaluation of humans and robots is possible. An experimental design is described and executed based on the developed game-based scenario, and exploratory analyses are performed based on the measured data
A Study on Sophisticated Production Management for Engineer-to-Order Production: A Mixed Integer Programming Formulation for Production Scheduling
No full textPart 1: Smart and Sustainable Supply Chain Management in the Society 5.0 EraInternational audienceEngineer-to-order (ETO) production in which products are designed and manufactured in response to customer orders is required to respond flexibly to customer requests at various stages from design to maintenance. This characteristic makes it difficult to apply a standard production planning strategy which divides the planning into three phases, i.e., long-term, medium-term and short-term planning (production scheduling), because there are large discrepancies among the phases and rescheduling requires a lot of man-hours. We proposed a production planning framework that unifies the granularity of resources and unit time in all of the planning phases aiming to reduce the discrepancies, and a model that is commonly used in the three phases of planning was organized as flexible job-shops. This paper provides a mixed integer programming formulation of the production scheduling problem based on the model considering the following characteristics of the target ETO production site: (1) The planner has discretion in shortening required processing time; (2) Operation time is limited to day time of weekdays; (3) Overtime works can be accepted if necessary; (4) Some operations of multiple parts must be processed at the same time on the same machine. A numerical experiment showed validity of the model
Designing Augmented Reality Assistance Systems for Operator 5.0 Solutions in Assembly
No full textPart 2: Human-centred Manufacturing and Logistics Systems Design and Management for the Operator 5.0International audienceIndustry 5.0 emphasises how technology may benefit humans and marks a move towards a socio-technical paradigm. This study looks at how Augmented Reality (AR) can be integrated into human-centered smart manufacturing systems to improve operator performance, especially when it comes to assembly and disassembly work. Relevant AR applications in manufacturing are found through a methodical assessment of the literature, emphasising the necessity of human-centered design methodologies. The paper then offers basic recommendations for integrating AR systems into manual workstations in an efficient manner with the goal of enhancing operator productivity and welfare. The background, motivation and methods are discussed. The main findings include specific considerations for supporting the AR design in assembly, discussing the relevance of targeting group of users, choicing the suitable devices according to the usability and developing effective instructions
Impact of Collaborative Robots on Human Trust, Anxiety, and Workload: Experiment Findings
No full textPart 4: Evolving Workforce Skills and Competencies for Industry 5.0International audienceThis work proposes an experiment setup and its protocols to investigate the impact of cobot’s size, speed and collaboration modes on different human factors including trust, propensity to trust, anxiety, and mental workload. The setup and the protocols supported the execution of different experiments where the 29 participants were asked to complete the Tower of Hanoi in collaboration with a cobot. The setup and the protocols provide a ready-to-use solution to expand experiments for further studies. Moreover, statistical analysis of the results shows higher cobot speeds increased trust propensity despite not significantly affecting overall trust or anxiety. Collaboration modes significantly influenced perceived workload and task performance, with the “Collaboration with Trigger” mode resulting in lower mental workload but longer task completion times. No significant differences were found in human factors concerning cobot size, indicating that variations in size do not significantly impact trust, propensity to trust, anxiety, or workload. Additionally, the collaboration mode with cobots notably affects workload perception and task performance, with specific modes reducing perceived effort but not necessarily improving task efficiency
Integrating Deep Learning Frameworks for Automated Medical Image Diagnosis
No full textPart 1: SDG 3 Good Health and Well-BeingInternational audienceIn today's healthcare system, medical image analysis is essential for diagnosis, treatment planning, and condition monitoring. In this study, the four well-known tools in the field are 3D Slicer, MONAI, SAMM, and YOLOv8 are examined and contrasted. The study's scope includes its ability to process a variety of 2D and 4D medical imaging datasets, which contain images, segmentation, annotations, and transformations. The process entails a comprehensive prehensive examination of every tool's feature, with particular emphasis on their capacity to read and write DICOM images, accommodate multiple file formats, offer interactive 3D Slicer. Key findings show that 3D Slicer performs exceptionally well when using deep learning techniques for segmentation and interactive visualization. While SAM Model exhibits adaptability in managing diverse segmentation prompts, MONAI provides a comprehensive end-to-end medical data processing solution. YOLOv8 shows potential for effective object detection. The study's implications include the possibility of using these instruments in clinical practices and medical research contexts. The research results provide insight into the current discussion about how to strategically integrate state-of-the-art methods to improve overall healthcare outcomes, diagnostic accuracy, and treatment planning efficiency as medical image analysis continues to advance
Closing the Communication Divide: Enhancing Sign Language Recognition with Gesture-to-Text Conversion Through Computer Vision
No full textPart 2: SDG 4 Quality EducationInternational audienceThis paper introduces an innovative method specifically created for identifying and understanding fixed hand movements that represent symbols in American Sign Language (ASL). With the goal of improving communication and learning for those with hearing and speaking difficulties, this system is exceptionally useful for the deaf community to interact with technology. Utilizing the widespread use of ASL as a widely accepted form of communication, the system primarily concentrates on recognizing and converting static hand gestures that correspond to ASL letters into written output. Furthermore, it also includes the ability to convert text to speech, adding to its functionality. At the heart of our groundbreaking approach is the use of Principal Component Analysis (PCA) on still images of the ASL alphabet to accurately detect gestures. By utilizing the power of PCA to identify crucial elements, our system can effectively classify input images and provide accurate recognition of the corresponding ASL alphabet. This output is then presented as text, and can even be transformed into speech, making it a comprehensive tool for individuals with hearing and speech impairments to seamlessly communicate through computer technology. Significantly, this system eliminates the requirement for extra data collecting tools, making it more convenient for users. The strong recognition of ASL gestures, along with the ability to convert those gestures into text and sound, empowers those who are deaf or hard of hearing to effectively communicate through technology
An NLP Based Approach to Automate and Enhance the Systematic Review Within PRISMA Format
No full textPart 2: SDG 4 Quality EducationInternational audienceSystematic Literature Review (SLR) is an integral part of research; however, to conduct an SLR, the manual review process would be time-consuming due to the high volume of literature involved. This paper presents a method that uses Natural Language Processing (NLP) and Machine Learning (ML) techniques to automate the systematic review process at two key stages: screening and eligibility assessment. The approach utilizes BERT (Bidirectional Encoder Representations from Transformers) embeddings and cosine similarity for automated label assignment and categorization of research articles. The proposed approach aligns to the principles of PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) in the automation of the domain based categorization of articles, and resolves the challenges posed by the state of art techniques. The proposed work considers three datasets with total 2164 records, collected based on identification criteria of PRISMA, to be classified into three categories. The proposed method based on BERT gives good results on categorization of article with an accuracy of 99% , 98% and 96% for each dataset