Local deep feature learning framework for 3D shape

Shuhui Bu, Pengcheng Han, Zhenbao Liu, Junwei Han, Hongwei Lin

Abstract: Highlights•We propose a novel framework to learn and extract local deep feature from several 3D shape descriptors.•The framework is not only limited to SI-HKS or AGD, other local descriptors are also supported.•The learning procedure is fully unsupervised.•There are no parameters to be turned in the learning procedure. Some other parameters have little influence on the performance, and it is easy to select proper parameters.