English  |  正體中文  |  简体中文  |  Items with full text/Total items : 888/888 (100%)
Visitors : 13062793      Online Users : 221
RC Version 7.0 © Powered By DSPACE, MIT. Enhanced by NTU Library IR team.
Scope Tips:
  • please add "double quotation mark" for query phrases to get precise results
  • please goto advance search for comprehansive author search
  • Adv. Search
    HomeLoginUploadHelpAboutAdminister Goto mobile version
    Please use this identifier to cite or link to this item: http://ccur.lib.ccu.edu.tw/handle/A095B0000Q/30

    Title: 以深度網路區域提案分類及回歸偵測漫畫中的文字區塊;Text Detection in Manga by Deep Region Proposal, Classification, and Regression
    Authors: 游智棋;YU, CHIH-CHI
    Contributors: 資訊工程研究所
    Keywords: 回歸網路;分類網路;區域提案網路;日本漫畫;文字偵測;空間轉換網路;Regression network;Classification network;Region proposal network;Manga;Text detection;Spatial transformer network
    Date: 2018
    Issue Date: 2019-05-23 10:30:13 (UTC+8)
    Publisher: 資訊工程研究所
    Abstract: 儘管在自然影像中的場景文字偵測(scene text detection)在學術研究中已經發展長遠,現存的自然影像文字偵測方法卻不適合直接應用在漫畫的文字偵測中,原因是漫畫中的文字有相當大的變異性與不同的上下文資訊。於本論文中,我們提出一個基於深層網路的漫畫文字偵測方法。此方法的主要偵測流程包含區域提案(region proposal),特徵提取(feature extraction),分類及回歸(classification/regression),我們將它們整合在單一的網路中。我們也結合一個空間轉換(spatial transformer)的網路以改善偵測的準確度。這個網路最主要的目的是藉由空間上扭曲轉變特徵,以達成能夠讓網路中的分類更加準確。空間轉換網路的整合讓我們的偵測方法能學習漫畫文字空間轉換上的特徵關係,將特徵加以轉換,最終使分類網路能夠正確分類困難或是變形劇烈的文字。最後經由實驗證明,我們的偵測方法表現優於現存的作法。因此在漫畫的文字偵測上,本論文提供一個目前為止最先進的方法。
    Though scene text detection for natural images has been studied for years, text in manga presents high variations and different contextual information, and existing scene text detection methods are not directly applicable. In this thesis, we propose an approach based on a deep network to detect text in manga. In this approach, primary processes of text detection, including region proposal, feature extraction, and classification/regression, are taken together in a single deep network. We also improve our approach by integrating a spatial transformer network. This network is proposed to deform the features spatially to make classification more accurate. In our work, the spatial transformer network is used to learn spatially deforming feature maps, and advances performance of the detector.The evaluation results show that our approach yields a big performance leap over the current state of the art, making it the leading method in manga text detection.
    Appears in Collections:[資訊工程學系] 學位論文

    Files in This Item:

    File Description SizeFormat

    All items in CCUR are protected by copyright, with all rights reserved.

    版權聲明 © 國立中正大學圖書館網頁內容著作權屬國立中正大學圖書館


    DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU Library IR team Copyright ©   - Feedback