Upright orientation of 3D shapes with Convolutional Networks

Liu, Zishun; Zhang, Jie; Liu, Ligang

doi:10.1016/j.gmod.2016.03.001

Cited by 26 publications

(34 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Besides chart images, there are several learning-based methods to recover the viewpoint and transfer function from a rendered image of 3D models and volumes. Liu et al [19] described a data-driven method for 3D model upright orientation estimation using a 3D CNN. Similarly, Kim et al [16] applied one CNN on 3D voxel data to generate a CNN shape feature for the upright orientation determination, and the other CNN to encode category-speci c information learned from a large number of 2D images on the web for the salient viewpoint detection.…”

Section: Reverse Engineering Of Visualizationsmentioning

confidence: 99%

CNNs Based Viewpoint Estimation for Volume Visualization

Shi

Tao

2019

ACM Trans. Intell. Syst. Technol.

View full text Add to dashboard Cite

Viewpoint estimation from 2D rendered images is helpful in understanding how users select viewpoints for volume visualization and guiding users to select be er viewpoints based on previous visualizations. In this paper, we propose a viewpoint estimation method based on Convolutional Neural Networks (CNNs) for volume visualization. We rst design an over t-resistant image rendering pipeline to generate the training images with accurate viewpoint annotations, and then train a category-speci c viewpoint classi cation network to estimate the viewpoint for the given rendered image. Our method can achieve good performance on images rendered with di erent transfer functions and rendering parameters in several categories. We apply our model to recover the viewpoints of the rendered images in publications, and show how experts look at volumes. We also introduce a CNN feature-based image similarity measure for similarity voting based viewpoint selection, which can suggest semantically meaningful optimal viewpoints for di erent volumes and transfer functions.

show abstract

Section: Reverse Engineering Of Visualizationsmentioning

confidence: 99%

CNNs Based Viewpoint Estimation for Volume Visualization

Shi

Tao

2019

ACM Trans. Intell. Syst. Technol.

View full text Add to dashboard Cite

show abstract

“…We first assume that the given 3D shapes are upright oriented, since most models in modern online repositories like ShapeNet [3], satisfy such requirements. For others do not have consistent orientation, we suggest the approach of [47] to deal with them.…”

Section: A Multi-view Representationmentioning

confidence: 99%

MVPN: Multi-View Prototype Network for 3D Shape Recognition

Yang

Wang

2019

IEEE Access

View full text Add to dashboard Cite

Convolutional Neural Networks (CNNs) have been widely used for 3D shape recognition. Despite its significant performance, we point out that there is still some rooms for improvement, including the lack of sufficient training samples and the extraction of discriminative features. In this work, we address the above limitations by suggesting a novel end-to-end learning framework named Multi-Veiw Prototype Network (MVPN). MVPN can jointly learn the features from multiple views of 3D shapes and the prototype representations of each class, recognition can then be performed by finding the nearest class prototype in the embedding space for a query sample. Furthermore, to obtain more discriminative features, we propose a discriminative loss, which encourages intra-class compactness and inter-class separability between learned representations, making the representation more discriminative and robust. Extensive experiments are conducted on two benchmarks: ModelNet dataset and ShapeNet Core55 dataset, and superior results have been achieved compared with state-of-the-art approaches.INDEX TERMS 3D shape recognition, prototype network, discriminative feature learning, convolutional neural networks.

show abstract

“…Complex preprocessing is likely to result in the deletion of important data. Additionally, the CNN network structure exhibits high tolerance for transformation (Arevalo et al., ), translation (Li et al., ), scaling (Li et al., ), tilting (Liu et al., ), etc. This property is useful for analyzing complex 3D scan data.…”

Section: Introductionmentioning

confidence: 99%

Convolutional Neural Network for Asphalt Pavement Surface Texture Analysis

Zheng

Gao

Sha

et al. 2018

Computer aided Civil Eng

View full text Add to dashboard Cite

Several data processing techniques (DPTs) have been implemented for evaluating pavement surface texture to partially replace onsite inspections by humans. However, the extensively varying real‐world situations have resulted in challenges in the widespread adoption of DPTs. To overcome these challenges, we propose the use of a convolutional neural network (CNN) to calculate the mean texture depth (MTD) without computing the surface texture feature statistics. Because a CNN is capable of automatically learning data features, the proposed method does not require the conjugation of DPTs for extracting features. The proposed CNN was trained and tested using 8,000 and 1,000 3D scan data samples, respectively, and achieved an average error of 0.0024 cm. The stability of the CNN was analyzed based on various test results. Comparative studies were conducted to verify the superiority of the CNN over conventional MTD algorithms. The results demonstrated that the CNN‐based method is significantly more stable in various real‐world situations. Additionally, the CNN‐based method achieved higher accuracy of automatic feature extraction than traditional MTD methods. Finally, the CNN‐based method was applied to evaluate the surface texture statistics of four highways in Shanxi, China, which were different from the training and testing samples; the results establish the transferability of this method to different highways.

show abstract

Upright orientation of 3D shapes with Convolutional Networks

Cited by 26 publications

References 31 publications

CNNs Based Viewpoint Estimation for Volume Visualization

CNNs Based Viewpoint Estimation for Volume Visualization

MVPN: Multi-View Prototype Network for 3D Shape Recognition

Convolutional Neural Network for Asphalt Pavement Surface Texture Analysis

Contact Info

Product

Resources

About