Enhanced Hand Gesture Recognition Using Image Transformer Model and Particle Swarm Optimization
Sonu Sonu, Nisha, Satya Narayan, Vinesh Kumar Jain · 2024
This study introduces the Image Transformer Model, a novel approach to enhance hand gesture recognition, leveraging recent advancements in deep learning and computer vision. By utilizing the Transformer's attention mechanism, the model excels in capturing intricate spatial correlations within visual data. It processes input as flattened patches, enabling comprehensive interpretation of spatial relationships in hand gesture images. The model thoughtfully combines various elements, including Multi-Head Self-Attention, FFN Network, Layer Normalization, & residual connections, resulting in superior performance compared to established models. The research further extends the model's capabilities through the incorporation of a Multi-Head Attention-based Transformer Network, complemented by Particle Swarm Optimization (PSO). By successfully optimizing hyperparameters with PSO, the model demonstrates improved recognition of complex spatial correlations in hand gesture images, outperforming such methods. The fusion of the Image Transformer Model & PSO tuning represents a significant advancement in the field of hand gesture identification, promising valuable applications across diverse domains.