Procedural Synthesis of Remote Sensing Images for Robust Change Detection with Neural Networks

Maria Kolos, Anton Marin, Alexey Artemov, Evgeny Burnaev

    Результат исследований: Глава в книге, отчете, сборнике статейМатериалы для конференциирецензирование

    4 Цитирования (Scopus)


    Data-driven methods such as convolutional neural networks (CNNs) are known to deliver state-of-the-art performance on image recognition tasks when the training data are abundant. However, in some instances, such as change detection in remote sensing images, annotated data cannot be obtained in sufficient quantities. In this work, we propose a simple and efficient method for creating realistic targeted synthetic datasets in the remote sensing domain, leveraging the opportunities offered by game development engines. We provide a description of the pipeline for procedural geometry generation and rendering as well as an evaluation of the efficiency of produced datasets in a change detection scenario. Our evaluations demonstrate that our pipeline helps to improve the performance and convergence of deep learning models when the amount of real-world data is severely limited.

    Язык оригиналаАнглийский
    Название основной публикацииAdvances in Neural Networks – ISNN 2019 - 16th International Symposium on Neural Networks, ISNN 2019, Proceedings
    РедакторыHuchuan Lu, Huajin Tang, Zhanshan Wang
    ИздательSpringer Verlag
    Число страниц17
    ISBN (печатное издание)9783030228071
    СостояниеОпубликовано - 2019
    Событие16th International Symposium on Neural Networks, ISNN 2019 - Moscow, Российская Федерация
    Продолжительность: 10 июл. 201912 июл. 2019

    Серия публикаций

    НазваниеLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
    Том11555 LNCS
    ISSN (печатное издание)0302-9743
    ISSN (электронное издание)1611-3349


    Конференция16th International Symposium on Neural Networks, ISNN 2019
    Страна/TерриторияРоссийская Федерация


    Подробные сведения о темах исследования «Procedural Synthesis of Remote Sensing Images for Robust Change Detection with Neural Networks». Вместе они формируют уникальный семантический отпечаток (fingerprint).