Hand Gesture Recognition and Cursor Control

Abstract

Abstract: Convolutional neural network in this process to overcome the issue, use the network AutoGesNet for the gesture recognition task that creating effectiveneural network architecture is challenging. To be more precise, we first combine and preprocess three sets of gesture recognition data. The AutoGesNet search space and general architecture are then designed. Additionally, we employ transfer learning and reinforcement learning techniques to automatically create the intricate AutoGesNet architecture. Finally, the searched neural network is adjusted and retrained for two alternative input sizes. The retrained model performs accurately on both our data set and the NUS Hand Posture Dataset II, according to experiments. network that performs wellin terms of recognition accuracy. We will contrast andmerge AutoGesNet in further work