An Improved 3D Shape Recognition Method Based on Panoramic View

Zheng, Qiang; Sun, Jian; Zhang, Le; Chen, Wei; Fan, Huanhuan

doi:10.1155/2018/6467957

Cited by 5 publications

(3 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There are five convolution layers in the 3V-DepthPano which use the Alexnet model [27]. Both Panoramic View [3] and DeepPano use four convolution Blocks with one convolution layer for each block, followed by a max-pooling layer, compared with four convolution layers in our method. The number of parameters in the fully-connected layers is not compared because there is no information on the number of feature vectors of this layer in the DeepPano method in [17].…”

Section: Resultsmentioning

confidence: 99%

“…It seems hard to select deep learning techniques directly for 3D shape recognition. Unlike the general data formation of 2D images, the extraordinary formation of 3D shapes, such as Point Clouds and meshes as direct inputs for CNN, is one of the most challenging in CNN generalization from 3D shapes to 2D images [3]. Presentations obtained for 3D shapes have significantly affected the performance of shape recognition.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A 3D Shape Recognition Method Using Hybrid Deep Learning Network CNN–SVM

2020

View full text Add to dashboard Cite

3D shape recognition becomes necessary due to the popularity of 3D data resources. This paper aims to introduce the new method, hybrid deep learning network convolution neural network-support vector machine (CNN-SVM), for 3D recognition. The vertices of the 3D mesh are interpolated to be converted into Point Clouds; those Point Clouds are rotated for 3D data augmentation. We obtain and store the 2D projection of this 3D augmentation data in a 32 × 32 × 12 matrix, the input data of CNN-SVM. An eight-layer CNN is used as the algorithm for feature extraction, then SVM is applied for classifying feature extraction. Two big datasets, ModelNet40 and ModelNet10, of the 3D model are used for model validation. Based on our numerical experimental results, CNN-SVM is more accurate and efficient than other methods. The proposed method is 13.48% more accurate than the PointNet method in ModelNet10 and 8.5% more precise than 3D ShapeNets for ModelNet40. The proposed method works with both the 3D model in the augmented/virtual reality system and in the 3D Point Clouds, an output of the LIDAR sensor in autonomously driving cars.Electronics 2020, 9, 649 2 of 14 connected information [4]. Hence, the right choice for the classification task is directly generating 3D Point Clouds for the original 3D shape.3D shape recognition is a base step that widely uses other tasks in intelligent electronic systems, such as 3D object tracking in intelligent robots or 3D object detection in autonomously driving cars. This paper will present the hybrid deep learning method, a combination of CNN and a Polynomial Kernel-based support vector machine (SVM) classifier, with a high accuracy in 3D shape recognition. CNN is used as the algorithm for feature extraction. The related studies are presented in section two before describing the method in section three. Then, we will compare other methods based on the numerical results. Finally, the conclusion is given. Related StudiesThere are two approaches: hand-crafted shape descriptors and the CNN-based method in some existing related methods for the recognition of 3D shapes. Descriptors of Hand-Crafted ShapeFeatures of a hand-crafted shape consist of local and global features [5]. Global shape features, for example, viewpoint histogram [6] and shape distributions [7], proceed with the whole shape but are inappropriate for the occluded shapes recognition of messy scenes. In contrast, 3D local shape features, for example, spin image [8], rotational projection statistics (RoPS) [9], heat kernel signatures (HKS) [10], and fast point feature histogram (FPFH) [11], or 2D image features extensions, 3D SURF [12] and 2.5D SIFT [13], outperform the global features in messy scenes. Various areas, including 3D shape matching, shape recognition, and 3D shape retrieval, have applied these methods successfully with heavy dependence on human design and field experience. As a result, working on the massive 3D repositories with various objects from a variety of domains is challenging for those shape features. CNN-Based Method

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A 3D Shape Recognition Method Using Hybrid Deep Learning Network CNN–SVM

2020

View full text Add to dashboard Cite

show abstract

“…PointNet [17] 77.60% N/A 3D ShapeNets [16] 83.54% 77.32% Geometry Image [18] 88.40% 83.90% DeepPano [19] 88.66% 82.54% PanoramicView [24] 89.80% 82.47% Our Method 90.20% 84.64%…”

Section: Algorithm Modelnet10 Modelnet40mentioning

confidence: 99%

A Deep Learning Method for 3D Object Classification Using the Wave Kernel Signature and A Center Point of the 3D-Triangle Mesh

et al. 2019

View full text Add to dashboard Cite

Computer vision recently has many applications such as smart cars, robot navigation, and computer-aided manufacturing. Object classification, in particular 3D classification, is a major part of computer vision. In this paper, we propose a novel method, wave kernel signature (WKS) and a center point (CP) method, which extracts color and distance features from a 3D model to tackle 3D object classification. The motivation of this idea is from the nature of human vision, which we tend to classify an object based on its color and size. Firstly, we find a center point of the mesh to define distance feature. Secondly, we calculate eigenvalues from the 3D mesh, and WKS values, respectively, to capture color feature. These features will be an input of a 2D convolution neural network (CNN) architecture. We use two large-scale 3D model datasets: ModelNet10 and ModelNet40 to evaluate the proposed method. Our experimental results show more accuracy and efficiency than other methods. The proposed method could apply for actual-world problems like autonomous driving and augmented/virtual reality.

show abstract