Gesture Assisted Real-Time Implementation of Rock Paper Scissors Employing SqueezeNet

One of the most widely used tasks in computer vision is Gesture Recognition, where state-of-the-art performance has been achieved using deep learning approaches. Applications that employ exclusive input of gesture for triggering tasks may have limitations in specific scenarios, such as dependent complex operations or multitasking tasks; however, with innovations in sensor technologies, improvement in response time, and enhanced accuracy of devices, hybrid voice-plus-gesture systems are finding utilization in diversified applications such as healthcare, gaming, surveillance systems, home automation, and industry applications. This has opened several verticals for research. Dataset for gesture assisted tasks is generally in the form of image/video, from which the information of moving objects is extracted, and redundant background data is ignored or removed. The work proposed describes how computer vision, OpenCV, and Python can be employed to create a virtual Rock Paper Scissors game supported by hand gesture recognition. Model was trained using the SqueezeNet architecture and had a test-data accuracy of 98%; utilizing modules like CVzone and Pygame, the game logic was developed in Python. Results obtained from the model developed successfully identified and categorized hand motions, offering a fun and captivating gaming experience. Results indicate that the implementation gave a real-time experience at par with existing gaming experience present online.