期刊名称：International Journal of Advances in Engineering and Management
电子版ISSN：2395-5252
出版年度：2020
卷号：2
期号：1
页码：880-885
语种：English
出版社：IJAEM JOURNAL
摘要：Sketch based image retrieval (SBIR) has received a lot of attention in the recent years. In this paper we propose an efficient SBIR system. The approach that has been employed here in SBIR system focuses on the sketches of the objects. Here, the user interface asks for a sketch image as an input from the user, on which the search has to be performed. Sketch based image retrieval is an approach where the system fetches images from the trained database, based on the query sketch received from the user through an interface. The images have embeddings which have to be extracted in order to build the data before processing further. We are making use of linear support vector machine (SVM). SVM is an algorithm that provides analysis of data for classification. We are using RESNET-50 which is a CNN (convolutional neural network) model which is constructed using pretrained images which is taken from the pickle model. A pre-trained VGG network is used as a input model and it finds the similarity between the sketch and image database. Tkinter is a graphical user interface used. Basically when we upload the image it will load to trained dataset first and then it will process it with the original images. With the help of pickle model it picks all the similar images present in the database. The system displays at most nine images as output, matching the input sketch.
关键词：Sketch based image retrieval;RESNET-50;Pre-trained VGG16;CNN
