SciTePress - Affiliation Profile

LIRIS-CNRS, France

Context (please use complete words only)

Field

Sort by

20 Papers

Adaptation in Learning Analytics Dashboards: A Systematic Review

Topics: Intelligent Learning and Teaching Systems; Adaptive Educational Systems; Learning Analytics and Educational Data Mining

Authors: Rémi Barbé ¹ , Benoît Encelle ¹ and Karim Sehaba ²

Affiliations: ¹ Univ. Lyon, UCBL, CNRS, INSA Lyon, Centrale Lyon, Univ. Lyon 2, LIRIS, UMR5205, F-69622 Villeurbanne, France , ² Univ. Lyon, Univ. Lyon 2, CNRS, INSA Lyon, UCBL, Centrale Lyon, LIRIS, UMR5205, F-69676 Bron, France

Digital Object Identifier: 10.5220/0012628600003693

Proceedings of CSEDU 2024, ISBN: 978-989-758-697-2

Quick Abstract

Although learning analytics dashboards (LAD) grow in numbers, they often fail to improve learner awareness as they lack adaptation capabilities. This paper presents a systematic review following the PRISMA statement, about the adaptation capabilities of LADs based on new definitions for LADs and learning indicators. A detailed analysis of 23 articles selected among 426 articles retrieved from databases was conducted based on a coding scheme, centered on adaptation and its dimensions, namely: to whom, what, to what, who, and how. The main result of this study is that there is more evidence of adaptable LADs than adaptive LADs. As a result, the road to adaptivity is worth exploring. The analysis of LAD’s common features led us to distinguish mainly 4 adaptable capabilities and 2 adaptive ones. Most of the adaptable capabilities consist of giving exploration power to the user and providing him with data filtering, zooming, or selection functionalities. In contrast, users have limited options when it comes to selecting indicators, their visualizations, and organization on the dashboard. Providing more flexible LADs could enhance their usability and increase learner awareness. Furthermore, the few adaptive features involve adaptations based on “if-then” rules and there are no reports of advanced computing techniques such as machine learning that could empower LAD’s adaptation.

Leveraging Unsupervised and Self-Supervised Learning for Video Anomaly Detection

Topics: Video Surveillance and Event Detection; Deep Learning for Visual Understanding

Authors: Devashish Lohani ^{1

;

2} , Carlos Crispim-Junior ¹ , Quentin Barthélemy ² , Sarah Bertrand ² , Lionel Robinault ^{1

;

2} and Laure Tougne Rodet ¹

Affiliations: ¹ Univ Lyon, Univ Lyon 2, CNRS, INSA Lyon, UCBL, LIRIS, UMR5205, F-69676 Bron, France , ² Foxstream, F-69120 Vaulx-en-Velin, France

Digital Object Identifier: 10.5220/0011663600003417

Proceedings of VISAPP 2023/VISIGRAPP, ISBN: 978-989-758-634-7

Quick Abstract

Video anomaly detection consists of detecting abnormal events in videos. Since abnormal events are rare, anomaly detection methods are mainly not fully supervised. One such popular family of methods learn normality by training an autoencoder (AE) on normal data and detect anomalies as they deviate from this normality. But the powerful reconstruction capacity of AE makes it still difficult to separate anomalies from normality. To address this issue, some works enhance the AE with an external memory bank or attention modules but still these methods suffer in detecting diverse spatial and temporal anomalies. In this work, we propose a method that leverages unsupervised and self-supervised learning on a single AE. The AE is trained in an end-to-end manner and jointly learns to discriminate anomalies using three chosen tasks: (i) unsupervised video clip reconstruction; (ii) unsupervised future frame prediction; (iii) self-supervised playback rate prediction. Furthermore, to correctly emphasize the detected anomalous regions in the video, we introduce a new error measure, called the blur pooled error. Our experiments reveal that the chosen tasks enrich the representational capability of the autoencoder to detect anomalous events in videos. Results demonstrate our approach outperforms the state-of-the-art methods on three public video anomaly datasets.

Synthetic Driver Image Generation for Human Pose-Related Tasks

Topics: Object Detection and Localization; Deep Learning for Visual Understanding ; Assistive Computer Vision

Authors: Romain Guesdon , Carlos Crispim-Junior and Laure Tougne Rodet

Affiliation: Univ. Lyon, Univ. Lyon 2, CNRS, INSA Lyon, UCBL, Centrale Lyon, LIRIS UMR5205, F-69676 Bron, France

Digital Object Identifier: 10.5220/0011780800003417

Proceedings of VISAPP 2023/VISIGRAPP, ISBN: 978-989-758-634-7

Quick Abstract

The interest in driver monitoring has grown recently, especially in the context of autonomous vehicles. However, the training of deep neural networks for computer vision requires more and more images with significant diversity, which does not match the reality of the field. This lack of data prevents networks to be properly trained for certain complex tasks such as human pose transfer which aims to produce an image of a person in a target pose from another image of the same person. To tackle this problem, we propose a new synthetic dataset for pose-related tasks. By using a straightforward pipeline to increase the variety between the images, we generate 200k images with a hundred human models in different cars, environments, lighting conditions, etc. We measure the quality of the images of our dataset and compare it with other datasets from the literature. We also train a network for human pose transfer in the synthetic domain using our dataset. Results show that our dataset matches the quality of existing datasets and that it can be used to properly train a network on a complex task. We make both the images with the pose annotations and the generation scripts publicly available.

Multitask Metamodel for Keypoint Visibility Prediction in Human Pose Estimation

Topics: Object Detection and Localization; Video Surveillance and Event Detection; Deep Learning for Visual Understanding ; Multi-task learning

Authors: Romain Guesdon , Carlos Crispim-Junior and Laure Tougne

Affiliation: Univ Lyon, Lyon 2, LIRIS UMR 5205 Lyon, France, F-69676

Digital Object Identifier: 10.5220/0010831200003124

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

Quick Abstract

The task of human pose estimation (HPE) aims to predict the coordinates of body keypoints in images. Even if nowadays, we achieve high performance on HPE, some difficulties remain to be fully overcome. For instance, a strong occlusion can deceive the methods and make them predict false-positive keypoints with high confidence. This can be problematic in applications that require reliable detection, such as posture analysis in car-safety applications. Despite this difficulty, actual HPE solutions are designed to always predict coordinates for each keypoint. To answer this problem, we propose a new metamodel that predicts both keypoints coordinates and their visibility. Visibility is an attribute that indicates if a keypoint is visible, non-visible, or not labeled. Our model is composed of three modules: the feature extraction, the coordinate estimation, and the visibility prediction modules. We study in this paper the performance of the visibility predictions and the impact of this task on the coordinate estimation. Baseline results are provided on the COCO dataset. Moreover, to measure the performance of this method in a more occluded context, we also use the driver dataset DriPE. Finally, we implement the proposed metamodel on several base models to demonstrate the general aspect of our metamodel.

Review of the Adaptability of a Set of Learning Games Meant for Teaching Computational Thinking or Programming in France

Topics: Serious Games; Gamification in Education; Game-based Learning; Educational Games; Programming Learning Tools

Authors: Hajar Saddoug ¹ , Aryan Rahimian ¹ , Bertrand Marne ² , Mathieu Muratet ³ , Karim Sehaba ⁴ and Sébastien Jolivet ⁵

Affiliations: ¹ Sorbonne Université, Place Jussieu, Paris, France , ² ICAR UMR 5191, Université Lumière Lyon 2, Parvis René Descartes, Lyon, France , ³ Sorbonne Université, CNRS, INS HEA, LIP6, F-75005 Paris, France , ⁴ LIRIS - Université Lumière Lyon 2, avenue Pierre Mendès-France, Lyon, France , ⁵ IUFE, Université de Genève, rue du Général-Dufour, Genève, Switzerland

Digital Object Identifier: 10.5220/0011126400003182

Proceedings of GonCPL 2022, ISBN: 978-989-758-562-3

Quick Abstract

Our work is part of a broader research project on how French teachers and trainers can appropriate learning games dedicated to computer science and programming. To foster this appropriation, we aim to implement a meta-design approach that favours instrumental genesis. In this article, we describe our review of a selection of learning games to identify whether they are suitable for this approach. We introduce a set of rather generic and reusable criteria to characterize the instrumentalization of serious games. With these criteria we thoroughly review 10 games among 48 selected. Thus, for each game selected, the identified criteria point out the availability of means and tools allowing teachers and trainers to understand the game, but they also assess adaptability of the latter in relation to pedagogical needs. Our results show that the adaptability of most of these 10 games remains weak and out of reach for many teachers and trainers. Indeed, none of the reviewed games were able to meet the requirements set out in the framework of the meta-design approach.

Toward a Meta-design Method for Learning Games

Topics: Authoring Tools and Content Development; Game-Based and Simulation-Based Learning; Tools for Educational Communication and Collaboration

Authors: Marne Bertrand ¹ , Muratet Mathieu ^{2

;

3} and Sehaba Karim ⁴

Affiliations: ¹ ICAR UMR 5191, Université Lumière Lyon 2, France , ² INS HEA, 92150 Suresnes, France , ³ Sorbonne Université, CNRS, LIP6, F-75005 Paris, France , ⁴ Université de Lyon, CNRS. Université Lyon 2, LIRIS, UMR5205, F-69676, France

Digital Object Identifier: 10.5220/0010530203700376

Proceedings of CSEDU 2021, ISBN: 978-989-758-502-9

Quick Abstract

To support the appropriation of serious learning games by teachers, we are studying design methods that focus on both design and use phases: meta-design methods. Our objective is to propose models and tools for designing levels and scenarios for Blockly Maze and to provide the teacher with monitoring indicators allowing him/her to appropriate and adapt the game according to the observed uses. In this article, we detail the first contributions of this study based on qualitative data: analysis of the game Blockly Maze and interviews with two teachers. First results draw three main needs: level design tool, scenario design tool and monitoring tool. Beyond these needs we introduce underlying models of each of them and future works.

CAD-based Learning for Egocentric Object Detection in Industrial Context

Topics: Object Detection and Localization; Mobile and Egocentric Object Detection and Recognition; First Person Vision

Authors: Julia Cohen ^{1

;

2} , Carlos Crispim-Junior ² , Céline Grange-Faivre ¹ and Laure Tougne ²

Affiliations: ¹ DEMS, Saint Bonnet de Mûre, France , ² Univ. Lyon, Lyon 2, LIRIS, F-69676 Lyon, France

Digital Object Identifier: 10.5220/0008975506440651

Proceedings of VISAPP 2020/VISIGRAPP, ISBN: 978-989-758-402-2

Quick Abstract

Industries nowadays have an increasing need of real-time and accurate vision-based algorithms. Although the performance of object detection methods improved a lot thanks to massive public datasets, instance detection in industrial context must be approached differently, since annotated images are usually unavailable or rare. In addition, when the video stream comes from a head-mounted camera, we observe a lot of movements and blurred frames altering the image content. For this purpose, we propose a framework to generate a dataset of egocentric synthetic images using only CAD models of the objects of interest. To evaluate different strategies exploiting synthetic and real images, we train a Convolutional Neural Network (CNN) for the task of object detection in egocentric images. Results show that training a CNN on synthetic images that reproduce the characteristics of egocentric vision may perform as well as training on a set of real images, reducing, if not removing, the need to manually annotate a large quantity of images to achieve an accurate performance.

Learner Performance Prediction Indicators based on Machine Learning

Topics: Machine Learning; Learning Analytics; Educational Data Mining

Author: Karim Sehaba

Affiliation: Université de Lyon, CNRS. Université Lyon 2, LIRIS, UMR5205, F-69676, France

Digital Object Identifier: 10.5220/0009396100470057

Proceedings of CSEDU 2020, ISBN: 978-989-758-417-6

Quick Abstract

This work is interested in the analysis of learners’ performances in order to define indicators to predict their results based on their interactions with a learning environment. These indicators should alert learners at risk, or their teachers, by highlighting their difficulties in order to help them get around them before it is too late. For this, we have defined a trace analysis approach based on the use of machine learning methods. This approach consists of preparing the plotted data automatically and manually, by selecting the attributes relevant to learning, then automatically extracting indicators explaining the learner’s results. Our work was applied to a data set resulting from a real training comprising 32593 learners producing 10 655 280 events. The accuracy of our predictions has reached around 80%. Rules extraction methods were also applied in order to explain the rules which govern the prediction indicator.

Efficient Bark Recognition in the Wild

Topics: Color and Texture Analyses; Features Extraction

Authors: Rémi Ratajczak ¹ , Sarah Bertrand ² , Carlos Crispim-Junior ² and Laure Tougne ²

Affiliations: ¹ Univ Lyon, Lyon 2, LIRIS, F-69676 Lyon, France, Unité Cancer et Environnement, Centre Léon Bérard, Lyon, France, Agence de l’Environnement et de la Maítrise de l’Energie, Angers and France , ² Univ Lyon, Lyon 2, LIRIS, F-69676 Lyon and France

Digital Object Identifier: 10.5220/0007361902400248

Proceedings of VISAPP 2019/VISIGRAPP, ISBN: 978-989-758-354-4

Quick Abstract

In this study, we propose to address the difficult task of bark recognition in the wild using computationally efficient and compact feature vectors. We introduce two novel generic methods to significantly reduce the dimensions of existing texture and color histograms with few losses in accuracy. Specifically, we propose a straightforward yet efficient way to compute Late Statistics from texture histograms and an approach to iteratively quantify the color space based on domain priors. We further combine the reduced histograms in a late fusion manner to benefit from both texture and color cues. Results outperform state-of-the-art methods by a large margin on four public datasets respectively composed of 6 bark classes (BarkTex, NewBarkTex), 11 bark classes (AFF) and 12 bark classes (Trunk12). In addition to these experiments, we propose a baseline study on Bark-101, a new challenging dataset including manually segmented images of 101 bark classes that we release publicly.

Related Ontology Subjects/Areas/Topics (4)

Color and Texture Analyses		Features Extraction		Image and Video Analysis
Computer Vision, Visualization and Computer Graphics

Generation of Task Models from Observed Usage Application to Web Browsing Assistance

Topics: Communication, Collaboration and Information Sharing; Tools and Technology for Knowledge Management; Knowledge Management Projects; Intelligent Information Systems; Information Technology and Information System Research

Authors: Karim Sehaba ¹ and Benoît Encelle ²

Affiliations: ¹ Université de Lyon, CNRS, Université de Lyon 2, LIRIS, UMR5205 and France , ² Université de Lyon, CNRS, Université de Lyon 1, LIRIS, UMR5205 and France

Digital Object Identifier: 10.5220/0008068300740082

Proceedings of KMIS 2019/IC3K, ISBN: 978-989-758-382-7

Quick Abstract

This work focuses on the extraction of knowledge from observed usage. More specifically, it aims to design a Web browsing assistance system that helps the user in carrying out his browsing task, or the designer in adapting or redesigning his Web application, according to real usage. The proposed approach consists of generating task models from interaction traces, which are then used to perform assistance. The characteristics to be supported by a task metamodel for assistance purposes are first identified and then confronted with the characteristics of existing task metamodels. This first study led us to choose the ConcurTaskTrees (CTT) metamodel. We then developed processes to generate CTT task models from traces. Finally, to validate our approach, we conducted unit testing and validation based on two real web browsing scenarios.

Related Ontology Subjects/Areas/Topics (8)

Artificial Intelligence	Knowledge Management and Information Sharing	Symbolic Systems
Communication, Collaboration and Information Sharing	Knowledge Management Projects	Tools and Technology for Knowledge Management
Intelligent Information Systems	Knowledge-Based Systems

Topic(s):

Title:
ISBN:
Year:
Acronym:
Subject:

Title:
Author:
Affiliation:
Subject:

Name:
Affiliation:
Country:
Conference:
Subject:

Name:
Country:
Subject:

Affiliation

Filter listed results, by