In many learning tasks, including semantic image segmentation, performance can be effectively improved through the fine-tuning of a pre-trained convolutional network, instead of training from scratch. With fine-tuning, the underlying assumption is that the pre-trained model extracts generic features, which are at least partially relevant for solving a segmentation task, but that would be difficult to extract from the smaller amount of data that is available for training in urban driving scenes segmentation. However, besides the initialization with the pre-trained model and the early stopping, there is no mechanism in classical fine-tuning approaches for keeping the generic features. Even worse, the standard weight decay drives the parameters towards the origin and affects the learned features. In this paper, we show that a simple regularization that uses the pre-trained model as a reference consistently improves the performance when applied to semantic urban driving scene segmentation. Experiments are done on the Cityscapes dataset, with four different architectures of convolutional networks.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    A Simple Weight Recall for Semantic Segmentation: Application to Urban Scenes


    Contributors:


    Publication date :

    2018-06-01


    Size :

    7581213 byte




    Type of media :

    Conference paper


    Type of material :

    Electronic Resource


    Language :

    English



    Real-Time High-Performance Semantic Image Segmentation of Urban Street Scenes

    Dong, Genshun / Yan, Yan / Shen, Chunhua et al. | IEEE | 2021


    CNN based semantic segmentation for urban traffic scenes using fisheye camera

    Deng, Liuyuan / Yang, Ming / Qian, Yeqiang et al. | IEEE | 2017


    Domain-Incremental Semantic Segmentation for Traffic Scenes

    Liu, Yazhou / Chen, Haoqi / Lasang, Pongsak et al. | IEEE | 2025


    Few-Shot Semantic Segmentation for Complex Driving Scenes

    Zhou, Jingxing / Chen, Ruei-Bo / Beyerer, Jurgen | IEEE | 2024