Web 3D Lightweight Learning Framework for Shape Retrieval Based on CNN

With the rapid development of Web3D technologies, it is more and more import and necessary that sketch-based model retrieval. Besides, 3D model displaying over web browser becomes important. In this paper, we propose simplificationbased lightweight method for shape to visualize over browser based on mobile Internet environment. Besides, a CNN-based learning method is conducted to obtain the best view of shape. Furthermore, learning framework is presented to conduct the final retrieval. Moreover, a feature fusion method is also used to generate a learning dictionary. In addition, proposed framework can provide new alternatives for shape retrieval in Web3D environment. What is more, the innovation is mainly presented by employing deep learning method to solve the best view of model cross-domains feature learning and fusion problems, based on mobile Internet retrieval results fast and convenient visualization problem. Last but not least, the experiment is realized to verify the feasibility of the framework. Especially, compared with many state-of-the-art mainstream approaches; the results show that the approach was superior.