Ensuring accessibility for individuals with cognitive impairments is essential for autonomy, self-determination, and full citizenship. However, manual Easy-to-Read (ETR) text adaptations are slow, costly, and difficult to scale, limiting access to crucial information in healthcare, education, and civic life. AI-driven ETR generation offers a scalable solution but faces key challenges, including dataset scarcity, domain adaptation, and balancing lightweight learning of Large Language Models (LLMs). In this paper, we introduce ETR-fr, the first dataset for ETR text generation fully compliant with European ETR guidelines. We implement parameter-efficient fine-tuning on PLMs and LLMs to establish generative baselines. To ensure high-quality and accessible outputs, we introduce an evaluation framework based on automatic metrics supplemented by human assessments. The latter is conducted using a 36-question evaluation form that is aligned with the guidelines. Overall results show that PLMs perform comparably to LLMs and adapt effectively to out-of-domain texts. Code and datasets are available at https://github.com/FrLdy/ETR-fr.

Simplifying complex texts is essential to ensure equitable access to information, particularly for individuals with cognitive impairments.The Easy-to-Read (ETR) initiative provides a framework to make content more accessible for these individuals.However, manually creating such texts remains time-consuming and resourceintensive.In this work, we investigate the potential of large language models (LLMs) to automate the generation of ETR content.To address the scarcity of aligned corpora and the specific constraints of ETR, we propose a multitask learning (MTL) approach that trains models jointly on text summarization, text simplification, and ETR generation.We explore two complementary strategies: multi-task retrievalaugmented generation (RAG) for in-context learning (ICL), and MTL-LoRA for parameterefficient fine-tuning (PEFT).Our experiments with Mistral-7B and LLaMA-3-8B, conducted on ETR-fr, a new high-quality dataset, show that MTL-LoRA consistently outperforms all other strategies in in-domain settings, while the MTL-RAG-based approach achieves better generalization in out-of-domain scenarios.

As Picasso said, a painting lives only through the one who looks at it. To materialize this thought, we propose to automatically produce artworks that visually transform paintings by amplifying and distorting the most observed areas by viewers. Our work is based on a study conducted at the Caen Museum of Fine Arts in France. During the study, 151 participants were equipped with eye-tracking glasses, and observed various paintings, first alone and then in pairs. Based on the fixation and gaze path stored data, we first generate saliency maps that reflect the visual attention given to each painting. These maps are then used to fine-tune the UNETRSal model, a neural network designed to predict saliency maps, in order to align its outputs with human visual patterns observed during the experiment. The saliency maps generated are subsequently used to create deformations of the original painting. This overall process gives rise to a new artwork born from the interaction between human gaze and AI-prediction.

Investigating realistic visual exploration is quite challenging in sport climbing, but it promises a deeper understanding of how performers adjust their perception-action couplings during task completion. However, the samples of participants and the number of trials analyzed in such experiments are often reduced to a minimum because of the time-consuming treatments of the eye-tracking data. Notably, mapping successive points of gaze from local views to the global scene is generally performed manually by watching eye-tracking video data frame by frame. This manual procedure is not suitable for processing a large number of datasets. Consequently, this study developed an automatic method for solving this global point of gaze localization in indoor sport climbing. Particularly, an eye-tracking device was used for acquiring local image frames and points of gaze from a climber’s local views. Artificial landmarks, designed as four-color-disk groups, were distributed on the wall to facilitate localization. Global points of gaze were computed based on planar homography transforms between the local and global positions of the detected landmarks. Thirty climbing trials were recorded and processed by the proposed methods. The success rates (Mean[Formula: see text]±[Formula: see text]SD) were up to 85.72%[Formula: see text]±[Formula: see text]13.90%, and the errors (Mean[Formula: see text]±[Formula: see text]SD) were up to [Formula: see text][Formula: see text]m. The proposed method will be employed for computing global points of gaze in our current climbing dataset for understanding the dynamics intertwining of gaze and motor behaviors during the climbs.

Action recognition is actually considered as one of the most challenging areas in computer vision domain. In this paper, we propose a new approach based on utilization of motion boundaries to generate Motion Stable Shape (MSS) features to describe human actions in videos. In fact, we have considered actions as a set of human poses. Temporal evolution of each human pose is modeled by a set of new MSS feature's. Motion stable shapes of considered poses are defined by specific regions located at the borders of movements. Our modelisation is composed of different steps. First, a volume of optical flow frames highlighting the principal motions in poses is substracted. Then, motion boundaries are computed from the previous optical flow frames. Finally, maximally Stable Extremal Regions (MSER) are applied to motion boundaries frames in order to obtain MSS features. To predict classes of different human actions, the MSS features are combined with a standard bag-of-words representation. To prove the efficiency of our developed model, we have performed a set of experiments on four datasets: Weizmann, KTH, UFC and Hollywood. Obtained experimental results show that the proposed approach significantly outperforms state-of-the-art methods.

We present an inpainting method for images and videos based on nonlocal discrete p-Laplace regularization on weighted graphs. Our work has the advantage of unifying local geometric methods and nonlocal exemplar-based ones in the same framework. Our image inpainting benefits from local and nonlocal regularities within the image. In addition to that, our video inpainting exploits temporal and spatial redundancies in order to obtain high quality results by considering a video sequence as a volume and not as a sequence of still frames. However, our method does not employ any motion estimation for video inpainting. Experiments demonstrate that our nonlocal method outperforms the local one by completing missing data with finer and more consistent details for textured and non-textured images and videos.

We present local and nonlocal algorithms for video denoising based on discrete regularization on graphs. The main difference between video and image denoising is the temporal redundancy in video sequences. Recent works in the literature showed that motion compensation is counter-productive for video denoising. Our algorithms do not require any motion estimation. In this paper, we consider a video sequence as a volume and not as a sequence of frames. Hence, we combine the contribution of temporal and spatial redundancies in order to obtain high quality results for videos. To enhance the denoising quality, we develop a nonlocal method that benefits from local and nonlocal regularities within the video. Experiments show that the nonlocal method outperforms the local one by preserving finer details at the expense of an increase in the computational effort. We propose an optimized method that is faster than the nonlocal approach, while producing equally attractive results.

Along with the ever-growing Web comes the proliferation of objectionable content, such as sex, violence, racism, etc. We need efficient tools for classifying and filtering undesirable Web content. In this paper, we investigate this problem and describe WebGuard, an automatic machine learning-based pornographic Web site classification and filtering system. Unlike most commercial filtering products, which are mainly based on textual content-based analysis such as indicative keywords detection or manually collected black list checking, WebGuard relies on several major data mining techniques associated with textual, structural content-based analysis, and skin color related visual content-based analysis as well. Experiments conducted on a testbed of 400 Web sites including 200 adult sites and 200 nonpornographic ones showed WebGuard's filtering effectiveness, reaching a 97.4 percent classification accuracy rate when textual and structural content-based analysis was combined with visual content-based analysis. Further experiments on a black list of 12,311 adult Web sites manually collected and classified by the French Ministry of Education showed that WebGuard scored a 95.62 percent classification accuracy rate. The basic framework of WebGuard can apply to other categorization problems of Web sites which combine, as most of them do today, textual and visual content.

@inproceedings{CI-CHAHIR-2003, author = {Hammami, M. and Chahir, Y. and Chen, L.}, title = {WebGuard: Web-Based Adult Content Detection and Filtering System}, booktitle = {IEEE/WIC International Conference on Web Intelligence (WI'03)}, pages = {574-578}, year = {2003}, address = {Halifax, Canada}, month = {October} }