English  |  正體中文  |  简体中文  |  Items with full text/Total items : 889/889 (100%)
Visitors : 14659654      Online Users : 18
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/33

    Title: 應用多向神經網路於多圖片多標籤分類;Multi-Stream Networks for Multi-Sampled Multi-Label Image Classification
    Authors: 郭紘睿;GUO, HUNG-JUI
    Contributors: 資訊工程研究所
    Keywords: 多標籤分類問題;深度神經網路;服裝類別分類;電影類別分類;服裝圖片;電影海報;多向神經網路;deep neural network;clothing images classification;movie genre classification;in?shop clothes images;Movie poster;multi-label classi?cation;multi-stream networks
    Date: 2018
    Issue Date: 2019-05-23 10:30:14 (UTC+8)
    Publisher: 資訊工程研究所
    Abstract: 在電腦視覺領域,圖片分類問題一直都是個熱門的議題。在本論文中我們提出一個深度神經網路,以實現電影海報和服裝的圖片分類。在此深度神經網路中,我們同時考慮視覺外觀及物件資訊以及考慮相同實體常以多張圖片呈現,將電影海報及服裝圖片依類別分類。由於一張圖可能屬於多個類別,我們將這個分類問題定義成一個多標籤分類問題。為了進行這項研究,我們蒐集了一個大型的電影海報資料集和一個服裝的資料集。基於這兩個資料集,我們訓練了一個卷積神經網路取得圖片的視覺資訊,並且使用現時最好的物件偵測方法以取得圖片的物件資訊。另外,我們整合相同實體的多張圖片呈現,提出多向深度神經網路進行整合。在最後的實驗中,我們達到比以往工作更好的分類效果。我們證明同時考慮相同實體中的多張圖片呈現,可以達到比只使用單張圖片更好的效果。
    In the computer vision research field, image classification has always been a hot topic. In order to achieve multi-label image classification for movie posters and clothing images, we propose a deep neural network in this thesis. In this network, we jointly consider visual appearance, object information, and multiple images of the same entity, and classify movie posters and clothing images. Since an image may belong to multiple categories, we define this classification problem as a multi-label classification problem. We collect a large-scale movie poster dataset and an in-shop clothes dataset, associated with various metadata. Based on these two datasets, we fine-tune a pretrained convolutional neural network to extract visual representation, and adopt a state-of-the-art framework to detect objects in images. In addition, we integrate multiple images of the same entity and propose a multi-stream deep neural network. In the evaluation, we show that the proposed method yields encouraging performance, which is much better than previous works. We also prove that jointly considering multiple samples of the same entity yields performance better than considering only one sample.
    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