Deep Learning-Based Action Recognition

• Autore: Lee Hyo Jong
• Pubblicazione: MDPI - Multidisciplinary Digital Publishing Institute, 2022
• Descrizione fisica: 1 electronic resource (240 p.)
• Soggetto topico: Technology: general issues; History of engineering & technology
• Soggetto non controllato: human action recognition; graph convolution; high-order feature; spatio-temporal feature; feature fusion; dynamic gesture recognition; multi-modalities network; class regularization; 3D-CNN; spatiotemporal activations; class-specific features; Dynamic Hand Gesture Recognition; human-computer interaction; hand shape features; pose estimation; stacked hourglass network; deep learning; convolutional receptive field; hand gesture recognition; human-machine interface; artificial intelligence; feedforward neural networks; spatio-temporal image formation; human activity recognition; fusion strategies; transfer learning; activity recognition; data augmentation; multi-person pose estimation; partitioned centerpose network; partition pose representation; continuous hand gesture recognition; gesture spotting; gesture classification; multi-modal features; 3D skeletal; CNN; spatiotemporal feature; embedded system; real-time; action recognition; Long Short-Term Memory; spatio-temporal differential
• Persona (resp. second.): LeeHyo Jong
• Sommario/riassunto: The classification of human action or behavior patterns is very important for analyzing situations in the field and maintaining social safety. This book focuses on recent research findings on recognizing human action patterns. Technology for the recognition of human action pattern includes the processing technology of human behavior data for learning, technology of expressing feature values ​​of images, technology of extracting spatiotemporal information of images, technology of recognizing human posture, and technology of gesture recognition. Research on these technologies has recently been conducted using general deep learning network modeling of artificial intelligence technology, and excellent research results have been included in this edition.
• Titolo autorizzato: Deep Learning-Based Action Recognition
• ISBN: 3-0365-5200-6
• Formato: Materiale a stampa
• Livello bibliografico: Monografia
• Lingua di pubblicazione: Inglese
• Record Nr.: 9910619465803321
• Lo trovi qui: Univ. Federico II