Portail HAL de Télécom Paris
Not a member yet
14191 research outputs found
Sort by
Texo: Formula Recognition within 20M Parameters
No full textIn this paper we present Texo, a minimalist yet highperformance formula recognition model that contains only 20 million parameters. By attentive design, distillation and transfer of the vocabulary and the tokenizer, Texo achieves comparable performance to state-of-the-art models such as UniMERNet-T and PPFormulaNet-S, while reducing the model size by 80% and 65%, respectively. This enables real-time inference on consumer-grade hardware and even in-browser deployment. We also developed a web application to demonstrate the model capabilities and facilitate its usage for end users
Le réseau des Micro-Folies, un potentiel “patrimoine en réseau” ?
No full textCarnet Hypothèses "Musées, patrimoines et pouvoir symbolique. Enjeux (géo)politiques et territoriaux du Patrimoine", https://mppsgeo.hypotheses.org/271
How to catch prospective use? A comparative study of virtual environment movie and guided imaginary for projecting future users in a public space
No full textInternational audienceFocusing on the design of a public space, we propose a methodological contribution to integrate possible future users to studies in prospective ergonomics. Our objective is to establish a rigorous protocol for comparing two methods of projecting potential usage. First, a projection interview based on a virtual environment movie and second, a Guided Imaginary Projection interview. We sought to determine to what extent these two methods provide participants with the ability to live a "quasi-experience" (Hoarau & Cahour, 2021) of the space. This "quasi-experience" is measured by a questionnaire inspired by the concepts of presence and absorption and by the analysis of the embodiment level of the discourse produced during the interviews (Allinc et al., 2019), from embodied to general discourse. The results show that while the two methods produce similar results in terms of the level of projection experience, the virtual environment movie projection produces significantly more general discourse
Walking lanes / walking lines: Bodily alignments and passing through doorways
No full textInternational audienceThere are very few studies that analyse the role of artefacts as shaping joint locomotion in public places. By video-recording pedestrians passing through doorways in a mall, we have observed how openings and doors contribute to mobile formations such as walking lanes or files. Doors play a major part as a focus for common direction. Doors occasion a modification of speed and a re-arrangement of spatial proximity between pedestrians during the process of passing through. We argue that mobile formats such as walking together in public places are based on culturally-methodic dynamics of bodily orientation to others. They are also based on a conjoint orientation to apertures that afford entry spaces to doors through which pedestrians wish to pass. Physical-artefactual boundaries such as doors, sidewalks and lanes play a major role in shaping joint locomotion. We would like to focus on a particular case of locomotion driven by artefacts: the passing through doors shaped by serial arrangements of pedestrians in a following/followed format. We treat this case of mobile formation as a specific genuine form of aggregate in its own right, distinct from side-by-side walking and other forms of mobile file
LEVERAGING WHOLE SLIDE DIFFICULTY IN MULTIPLE INSTANCE LEARNING TO IMPROVE PROSTATE CANCER GRADING
No full textInternational audienceMultiple Instance Learning (MIL) has been widely applied in histopathology to classify Whole Slide Images (WSIs) with slide-level diagnoses. While the ground truth is established by expert pathologists, the slides can be difficult to diagnose for non-experts and lead to disagreements between the annotators. In this paper, we introduce the notion of Whole Slide Difficulty (WSD), based on the disagreement between an expert and a non-expert pathologist. We propose two different methods to leverage WSD, a multi-task approach and a weighted classification loss approach, and we apply them to Gleason grading of prostate cancer slides. Results show that integrating WSD during training consistently improves the classification performance across different feature encoders and MIL methods, particularly for higher Gleason grades (i.e., worse diagnosis)
Désordre muséal à Saadiyat ? Ou comment ce laboratoire à taille urbaine reconfigure le paysage muséal mondial ?
No full textMusées, patrimoines et pouvoir symbolique. Enjeux (géo)politiques et territoriaux du Patrimoine, https://mppsgeo.hypotheses.org/2936Le projet de l’île Saadiyat à Abu Dhabi – soit la construction ambitieuse d’une île aux musées – a commencé il y a une vingtaine d’années. Diversement discuté en aménagement du territoire, en géopolitique et en études muséales, ce projet interroge sur ce qu’il apporte aux musées et à leur relation au territoire. En effet, par son ampleur, ses échecs et ses transformations, de même que par sa conception unifiée comme un laboratoire-musée des musées (et non pas comme un regroupement progressif d’institutions mitoyennes), il apparaît clairement que Saadiyat provoque un chamboulement dans le paysage muséal mondial. En y instaurant son nouvel ordre, le projet a pour effet de provoquer d’autres désordres - qui se répercutent à divers niveaux sur le(s) musée(s), tant ceux de l’île que ceux du reste du monde. Ainsi, comment le projet de Saadiyat transforme cette ville en véritable laboratoire et comment ce laboratoire muséal urbain chamboule le réseau mondial des musées en permettant le déploiement de voix périphériques dans la manière d’appréhender le musée ? Deux chamboulements se juxtaposent. Le premier est interne au territoire : de nombreux musées n’ont finalement jamais ouverts pour des raisons obscures, remodelant le paysage muséal de l’île. Le second a une effectivité qui lui est externe : l’implantation de ces musées bouscule le paysage muséal mondial en changeant les récits muséaux et offrant de nouvelles routes grâce à ce décentrement culturel
5G-EcoSim: A Simulation Framework for Estimating 5G Energy Consumption Using Real-World Data and Analytical Models
No full textInternational audienceThe definition and deployment of next-generation mobile networks must incorporate considerations of sustainability and environmental impact. In this context, estimating the energy consumption of a mobile network deployment across a city or region is of utmost importance. Nationwide historical aggregate values are informative but do not allow to understand the underlying dynamics or to perform prospective studies. There is hence a need for bottom-up approaches to assess the energy consumption of mobile networks. However, this is a complex task, as it depends on numerous factors, including the number and spatial distribution of base stations, the underlying technologies and their configuration, as well as user demand patterns and their geographic distribution. This paper thus introduces a simulation framework aimed at estimating the energy consumption of 5G networks at urban, regional or national scale. The framework integrates radio propagation models operating in the 3.5 GHz band with publicly available datasets describing user and base station locations, as well as network traffic volumes. As a case study, we consider the 5G deployment in France and examine the spatial distribution of network load and the resulting energy consumption. The source code and datasets are publicly available, ensuring that the simulator is fully reproducible and easily adaptable to other use cases, countries, or regions
S-PRESSO: Ultra Low Bitrate Sound Effect Compression With Diffusion Autoencoders And Offline Quantization
No full textInternational audienceNeural audio compression models have recently achieved extreme compression rates, enabling efficient latent generative modeling. Conversely, latent generative models have been applied to compression, pushing the limits of continuous and discrete approaches. However, existing methods remain constrained to low-resolution audio and degrade substantially at very low bitrates, where audible artifacts are prominent. In this paper, we present S-PRESSO, a 48kHz sound effect compression model that produces both continuous and discrete embeddings at ultra-low bitrates, down to 0.096 kbps, via offline quantization. Our model relies on a pretrained latent diffusion model to decode compressed audio embeddings learned by a latent encoder. Leveraging the generative priors of the diffusion decoder, we achieve extremely low frame rates, down to 1Hz (750x compression rate), producing convincing and realistic reconstructions at the cost of exact fidelity. Despite operating at high compression rates, we demonstrate that S-PRESSO outperforms both continuous and discrete baselines in audio quality, acoustic similarity and reconstruction metrics.</div
Craft-Based Data Physicalization: Opportunities and Challenges
No full textInternational audienceThis three-hour workshop will gather data visualization and HCIresearchers and practitioners to explore the possibilities of datarepresentation using craft techniques. Participants will submit a 2-4page document including (i) a statement of their craft experience,(ii) representative images of physicalizations they have createdusing this craft technique, and (iii) a discussion of opportunitiesand challenges for physicalizing data in their craft domain. Duringthe workshop, participants and organizers will work in groups tobrainstorm ways of representing data through their shared craft ofinterest. Then, every group proposes a synthesis of opportunitiesand challenges of the craft technique they worked with. Together,the community will chart a research agenda on how craft can ex-pand the design space of data physicalization, inform the creationof more expressive and accessible authoring tools, and raise newquestions around aesthetics, accuracy, and the role of slow makingin data representation.Cet atelier de trois heures réunira des chercheurs et des praticiens spécialisés dans la visualisation des données et l'IHM afin d'explorer les possibilités de représentation des données à l'aide de techniques artisanales. Les participants devront soumettre un document de 2 à 4 pages comprenant (i) une description de leur expérience artisanale, (ii) des images représentatives des physicalisations qu'ils ont créées à l'aide de cette technique artisanale, et (iii) une discussion sur les opportunités et les défis liés à la physicalisation des données dans leur domaine artisanal. Au cours de l'atelier, les participants et les organisateurs travailleront en groupes pour réfléchir à des moyens de représenter les données à travers leur artisanat commun. Ensuite, chaque groupe proposera une synthèse des opportunités et des défis liés à la technique artisanale sur laquelle il a travaillé. Ensemble, la communauté élaborera un programme de recherche sur la manière dont l'artisanat peut élargir l'espace de conception de la matérialisation des données, inspirer la création d'outils de création plus expressifs et plus accessibles, et soulever de nouvelles questions sur l'esthétique, la précision et le rôle de la fabrication lente dans la représentation des données