Deep multi-task networks are of particular interest for autonomous driving systems. They can potentially strike an excellent trade-off between predictive performance, hardware constraints and efficient use of information from multiple types of annotations and modalities. However, training such models is non-trivial and requires balancing learning over all tasks as their respective losses display different scales, ranges and dynamics across training. Multiple task weighting methods that adjust the losses in an adaptive way have been proposed recently on different datasets and combinations of tasks, making it difficult to compare them. In this work, we review and systematically evaluate nine task weighting strategies on common grounds on three automotive datasets (KITTI, Cityscapes and WoodScape). We then propose a novel method combining evolutionary meta-learning and task-based selective backpropagation, for computing task weights leading to reliable network training. Our method outperforms state-of-the-art methods by a significant margin on a two-task application.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    Dynamic Task Weighting Methods for Multi-task Networks in Autonomous Driving Systems


    Contributors:


    Publication date :

    2020-09-20


    Size :

    2359209 byte




    Type of media :

    Conference paper


    Type of material :

    Electronic Resource


    Language :

    English



    UMT-Net: A Uniform Multi-Task Network With Adaptive Task Weighting

    Chen, Sihan / Zheng, Lianqing / Huang, Libo et al. | IEEE | 2024


    Topology-aware Road Extraction via Multi-task Learning for Autonomous Driving

    Li, Tao / Ye, Shanding / Li, Ruihang et al. | IEEE | 2023


    Task Offloading for MEC-V2X Assisted Autonomous Driving

    Zhang, Yilun / Chen, Changrun / Zhu, Huiling et al. | IEEE | 2024


    Multi-Task Machine-Learned Models for Object Intention Determination in Autonomous Driving

    CASAS SERGIO / LUO WENJIE / URTASUN RAQUEL | European Patent Office | 2019

    Free access