Drawing Semantic Retrieval Algorithms Based on Deep Multilayer Convolutional Network

Abstract

Semantic retrieval based on painting has become a research hotspot in the field of pattern recognition and computer vision. Compared with traditional methods, the depth representation based on deep convolution neural network has obvious performance advantages in retrieval tasks. Therefore, this paper proposes a three-dimensional model retrieval method using hand-drawn image fusion information entropy and CNN. Firstly, the representative view of the model is obtained by semantic analysis of the drawing image, and the representative view is processed by edge detection to get the contour image. Then, the contour image and the sketch are input into CNN to extract the feature descriptor and match the features. Finally, this method is tested on SREC2013 database and the results show that its retrieval accuracy is higher than that of other traditional methods.