Search
3 articles for “Media Pipe model”
-
Sign Language and Face Expression Recognition Using Neural Networks: Deep Learning Approach to Break Communication Barriers
Abstract: Our study proposes a multimodal gesture recognition system specifically designed to aid communication for the deaf community. By employing neural network concepts, we utilize 3D convolutional neural networks (3D CNNs) to extract features from both hand and face images, focusing on relevant regions. Preprocessing techniques are applied to isolate these areas of interest prior to feature extraction. Unique 3D CNN architectures are then trained for each modality to capture the …
Published in Current Trends in Signal Processing · Vol. 14, Issue 3, 2024 · pp. 1–10 Read article
-
A Mathematical Perspective on Recent Cloud-Computing Trends for Scalable and Secure Social-Media Platforms
Abstract: Cloud computing underpins modern social-media platforms by providing elastic compute, storage, and data-processing pipelines capable of absorbing highly bursty workloads. This paper surveys recent cloud-native trends—serverless and event-driven design, container orchestration, edge/CDN offload, streaming analytics, and privacy-enhancing security controls—and formalizes their impact through a compact mathematical model. We express workload volatility using arrival-rate functions, use queueing-based capacity sizing to derive auto-scaling rules, and formulate an optimization objective that balances cost …
Published in Recent Trends in Mathematics · Vol. 3, Issue 2, 2026 · pp. 35–40 Read article
-
Machine Learning Optimization for VARTM Carbon Polymer Laminates
Abstract: Vacuum-assisted resin transfer moulding (VARTM) is a key low-cost, out-of-autoclave process for manufacturing large-scale carbon-fibre reinforced polymer (CFRP) laminates crucial to aerospace wings, wind-turbine blades, marine hulls, and automotive structures. Unpredictable resin flow often leads to voids, dry spots, and race-tracking defects, resulting in 27.9% scrap rates and lengthy, costly trial-and-error design cycles. Although surrogate models provide rapid impregnation predictions for simple flat-plate geometries, vision-based monitoring is limited to idealized …
Published in Journal of Polymer & Composites · Vol. 14, Issue 3, 2026 · pp. 229–245 Read article