Most 3D shape classification and retrieval algorithms were based on rigid 3D shapes, deploying these algorithms directly to nonrigid 3D shapes may lead to poor performance due to complexity and changeability of non-rigid 3D shapes. To address this challenge, we propose a fusion view convolutional neural networks (FVCNN) framework to extract the deep fusion features for non-rigid 3D shape classification and retrieval. We first propose a projection module to transform the nonrigid 3D shape into a
2022-09-08 23:41:05
3.62MB
研究论文
1