Optical Music Recognition (OMR) has made significant progress since its inception, with various approaches now capable of accurately transcribing music scores into digital formats. Despite these advancements, most so-called end-to-end OMR approaches still rely on multi-stage processing pipelines for transcribing full-page score images, which entails challenges such as the need for dedicated layout analysis and specific annotated data, thereby limiting the general applicability of such methods. In this paper, we present the first truly end-to-end approach for page-level OMR in complex layouts. Our system, which combines convolutional layers with autoregressive Transformers, processes an entire music score page and outputs a complete transcription in a music encoding format. This is made possible by both the architecture and the training procedure, which utilizes curriculum learning through incremental synthetic data generation. We evaluate the proposed system using pianoform corpora, which is one of the most complex sources in the OMR literature. This evaluation is conducted first in a controlled scenario with synthetic data, and subsequently against two real-world corpora of varying conditions. Our approach is compared with leading commercial OMR software. The results demonstrate that our system not only successfully transcribes full-page music scores but also outperforms the commercial tool in both zero-shot settings and after fine-tuning with the target domain, representing a significant contribution to the field of OMR.
In this paper we study singular limits of congestion-averse growth models, connecting different models describing the effect of congestion. These models arise in particular in the context of tissue growth. The main ingredient of our analysis is a family of energy evolution equations and their dissipation structures, which are novel and of independent interest. This strategy allows us to consider a larger family of pressure laws as well as proving the joint limit, from a compressible Brinkman's model to the incompressible Darcy's law, where the latter is a Hele-Shaw type free boundary problem.
Current research efforts are focused on replacing chemical compounds with high-value natural products. Accordingly, fucoidans which are a predominant bioactive compounds in seaweeds, showed promising biological and pharmacological potential. In this study, we aimed to extract, characterize and evaluate the antioxidant, anti-inflammatory and antinociceptive activities of fucoidan (Fuc-Sarg) from the brown seaweed Sargassum vulgare C. Agarth collected from Salammbo coast, Tunisia. Fuc-Sarg was isolated and characterized by different techniques as colorimetric and turbidimetric assays, Fourier-Transform Infrared spectroscopy (FTIR), 1H NMR spectroscopy, Size-Exclusion Chromatography (SEC) and Gas Chromatography-Mass Spectrometry (GC-MS) analysis. Results showed that Fuc-Sarg was obtained with an extraction yield of 3.34%, including 67.75% of total sugars, 21.10% of uronic acids, 13.5% sulfate groups and a low protein content (0.6%). FTIR and NMR analysis confirmed the presence of sulfated fucopyranose residues, while SEC showed a high molecular weight (Mw = 500,000 g/mol) with a dispersity of 13.8. Besides, GC-MS analysis revealed a heterogeneous monosaccharides composition including (arabinose, xylose, galactose, and L-fucose). However, the evaluation of the antioxidant potential of the fucoidan fraction was carried out using FRAP and DPPH assays. Results demonstrated a strong reducing power and free radical scavenging activity (IC50 = 21.2 mu g/mL). Pharmacological evaluation, in vivo, demonstrated significant dose-dependent anti-inflammatory effects in the xylene-induced ear edema model (up to 84.97% inhibition percentage) and interesting antinociceptive potential in hot plate and abdominal constriction writhing tests. These findings highlight the pharmacological therapeutic potential of fucoidan extracted from S. vulgare as a promising natural agent.
The relative efficacy of bimekizumab and risankizumab in patients with PsA who were biologic disease-modifying anti-rheumatic drug naïve (bDMARD naïve) or with previous inadequate response or intolerance to tumor necrosis factor inhibitors (TNFi-IR) was assessed at 52 weeks (Wk52) using matching-adjusted indirect comparisons (MAIC). Relevant trials were systematically identified. For patients who were bDMARD naïve, individual patient data (IPD) from BE OPTIMAL (NCT03895203; N = 431) were matched with summary data from KEEPsAKE-1 (NCT03675308; N = 483). For patients who were TNFi-IR, IPD from BE COMPLETE (NCT03896581; N = 267) were matched with summary data from the TNFi-IR patient subgroup in KEEPsAKE-2 (NCT03671148; N = 106). To adjust for cross-trial differences, patients from the bimekizumab trials were re-weighted to match the baseline characteristics of patients in the risankizumab trials. Adjustment variables were selected based on expert consensus (n = 5) and adherence to established MAIC guidelines. Recalculated bimekizumab Wk52 outcomes for American College of Rheumatology (ACR) 20/50/70 response criteria and minimal disease activity (MDA) index (non-responder imputation) were compared with risankizumab outcomes via non-placebo-adjusted comparisons. In patients who were bDMARD naïve, bimekizumab had a significantly greater likelihood of response than risankizumab at Wk52 for ACR50 (odds ratio [95
This study introduces a novel AI-based prediction framework for Fused Filament Fabrication (FFF) process optimization, integrating high-fidelity simulation with machine learning and design of experiments (DOE) analysis. In the proposed approach, a full-factorial DOE matrix is employed to drive Digimat-AM, a physics-based thermo-mechanical simulation tool, generating an exhaustive dataset to train six machine learning models (Random Forest, XGBoost, ANN, SVR, AdaBoost, and KNN). These models predict four critical responses: deflection, residual stress, print time, and shape tolerance (dimensional accuracy), and their performance is benchmarked against classical DOE response-surface models. The AI models exhibit outstanding predictive accuracy: ensemble methods (Random Forest, XGBoost) achieved near-perfect agreement with simulation outputs, significantly outperforming DOE models. Interpretability is ensured via 3D response-surface plots, which illustrate the influence of parameter interactions and align with known thermo-mechanical trends. This combination of predictive fidelity and transparent analysis delivers valuable design insights and practical utility. The trained ML models serve as fast surrogates (digital twins), enabling rapid “what-if” scenario exploration and reducing reliance on trial-and-error in print optimization. This AI-driven framework streamlines additive manufacturing workflows by accelerating process tuning and laying the foundation for intelligent, data-driven FFF optimization.
