Understanding the behaviors and intentions of humans are one of the main challenges autonomous ground vehicles still faced with. More specifically, when it comes to complex environments such as urban traffic scenes, inferring the intentions and actions of vulnerable road users such as pedestrians become even harder. In this paper, we address the problem of intent action prediction of pedestrians in urban traffic environments using only image sequences from a monocular RGB camera. We propose a real-time framework that can accurately detect, track and predict the intended actions of pedestrians based on a tracking-by-detection technique in conjunction with a novel spatio-temporal DenseNet model. We trained and evaluated our framework based on real data collected from urban traffic environments. Our framework has shown resilient and competitive results in comparison to other baseline approaches. Overall, we achieved an average precision score of 84.76% with a real-time performance at 20 FPS.


    Access

    Download


    Export, share and cite



    Title :

    Real-time Intent Prediction of Pedestrians for Autonomous Ground Vehicles via Spatio-Temporal DenseNet


    Contributors:

    Publication date :

    2019


    Remarks:

    Accepted to ICRA 2019


    Type of media :

    Preprint


    Type of material :

    Electronic Resource


    Language :

    English




    Intent Communication between Autonomous Vehicles and Pedestrians

    Matthews, Milecia / Chowdhary, Girish / Kieson, Emily | ArXiv | 2017

    Free access


    Impact of autonomous vehicles on pedestrians' safety

    Brar, Jaspreet Singh / Caulfield, Brian | IEEE | 2017



    Intent Prediction of Pedestrians via Motion Trajectories Using Stacked Recurrent Neural Networks

    Saleh, Khaled / Hossny, Mohammed / Nahavandi, Saeid | IEEE | 2018