Recognition of three-dimensional (3D) shape is a remarkable subject in computer vision systems, because of the lack of excellent\nshape representations. With the development of 2.5D depth sensors, shape recognition is becoming more important in practical\napplications.Many methods have been proposed to preprocess 3D shapes, in order to get available input data. A common approach\nemploys convolutional neural networks (CNNs), which have become a powerful tool to solve many problems in the field of\ncomputer vision. DeepPano, a variant of CNN, converts each 3D shape into a panoramic view and shows excellent performance. It\nis worth paying attention to the fact that both serious information loss and redundancy exist in the processing of Deep Pano, which\nlimits further improvement of its performance. In this work, we propose a more effective method to preprocess 3D shapes also based\non a panoramic view, similar to Deep Pano.We introduce a novel method to expand the training set and optimize the architecture\nof the network.The experimental results show that our approach outperforms Deep Pano and can deal with more complex 3D shape\nrecognition problems with a higher diversity of target orientation.
Loading....