Real-Time Hand Signal Detection Using Convolutional Neural Networks

This paper aims at developing a robust hand signal detection system using the popular deep learning framework of TensorFlow. By harnessing the power of convolutional neural networks (CNNs), the study endeavors to achieve real-time, accurate, and efficient recognition of hand signals from video or image data. The dataset serves as the foundation for training a CNN model using TensorFlow. Through the process of training, the model learns to extract discriminative features associated with different hand signals. Various pre-processing techniques, including data augmentation, normalization, and resizing are employed to enhance the model’s ability to generalize across different variations of hand signals. The developed hand signal detection system utilizes the trained CNN model for inference. Input video or image data is fed into the model, which analyzes and classifies hand signals based on the learned features. To enhance the real-time performance of the system, optimization techniques such as model quantization and hardware acceleration are explored and implemented. Performance metrics such as accuracy(96.5%), precision(0.91), recall(0.94), and F1-Score(0.92) are measured to assess the effectiveness and efficiency of the developed system and same results are declared in the result section.