文章基本信息

标题：Automatic Punjabi Caption Generation For Sports Images
本地全文：下载
作者：Manleen Kaur ; Gurpreet Josan ; Jagroop Kaur 等
期刊名称：INFOCOMP
印刷版ISSN：1807-4545
出版年度：2021
卷号：20
期号：1
页码：109-120
出版社：Federal University of Lavras
摘要：Image understanding and language generation have always been a difficult task in the field of Artificial Intelligence. Automatic Image Caption Generation is concerned with the task of understanding the image and generating a caption for it. In this paper, we represented our research work that uses the Deep Learning technique to create Punjabi captions for a given image and its associated news document. High-level features of the images are extracted using the pre-trained VGG-19 (Visual Geometry Group) model. These image features are merged with features of news text which are extracted using LSTM (Long Short Term Memory). The proposed model augments keywords from associated news text to generate suitable captions. Using both BLEU scores and human evaluations, we show that the proposed method is successful in generating intelligible and suitable captions.
其他摘要：Image understanding and language generation have always been a difficult task in the field of Artificial Intelligence. Automatic Image Caption Generation is concerned with the task of understanding the image and generating a caption for it. In this paper, we represented our research work that uses the Deep Learning technique to create Punjabi captions for a given image and its associated news document. High-level features of the images are extracted using the pre-trained VGG-19 (Visual Geometry Group) model. These image features are merged with features of news text which are extracted using LSTM (Long Short Term Memory). The proposed model augments keywords from associated news text to generate suitable captions. Using both BLEU scores and human evaluations, we show that the proposed method is successful in generating intelligible and suitable captions.
关键词：Image Caption; Deep Neural Network; Sequence to Sequence generation; Keyword Augmentation.