Research Of Large Scale Sketch Based Image Retrieval

We address the problem of large scale sketch based image retrieval. The search is based on content-based image retrieval. Users search the image by drawing a sketch. By comparing different method of edge detection, such as Canny edge detection, we chose the line segment detection method which has good performance in our paper. We take advantage of inverted index to improve the performance and struct tensor to improve the retrieval accuracy. In terms of system implementation, we make use of Hadoop in preprocess, and other engineering methods to optimize the retrieval process. Finally, we build a retrieval system for 3D cartoon creators searching 3D models and an app for commodities retrieval based on the research.