The application for this Traffic Police Hand Gesture Recognition system is to use Tensor Flow's MoveNet Thunder model in autonomous vehicle navigation. The system recognizes gestures commonly used by traffic police in India - ‘Stop’, ‘Turn Left’, ‘Move Forward’. We used a custom dataset of 8,000 images of gestures and environmental conditions. High accuracy is achieved with low over fitting using the architecture of neural network by using the dense and dropout layers. Indeed, Haar cascades for real time face detection make sure the gestures are processed only when the traffic officer is facing the camera. Minor misclassification is seen in some similar gestures and the model is accurate to 89%. Its robustness is underscored through validation through the Carla simulator, under different lighting and weather conditions. The approach put forward is boding well as a means to integrating gesture recognition into autonomous vehicles as a complement to safer driving in traffic-managed scenarios.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    Enhancing Autonomous Vehicle Navigation in Traffic Police Hand Gesture Recognition for Self Driving Cars in India Using MoveNet Thunder


    Contributors:


    Publication date :

    2025-01-07


    Size :

    626067 byte




    Type of media :

    Conference paper


    Type of material :

    Electronic Resource


    Language :

    English




    Enhanced BiLSTM for Traffic Police Gesture Recognition in Autonomous Vehicles

    Sunil, Vaidya Pranay / Anjali, A. / Devangan, Deepak Kumar | IEEE | 2024



    GESTURE RECOGNITION SYSTEM FOR AUTONOMOUS VEHICLE TRAFFIC CONTROL

    COHEN CHARLES J / JACOBUS CHARLES / BEACH GLENN et al. | European Patent Office | 2023

    Free access

    Traffic Police Gesture Recognition Based on Openpose and GRU

    Wang, Dengwen / Wang, Wangmeng / Chen, Yanbing et al. | IEEE | 2022