Hybrid CNNs: A Rotation Equivariant Framework for High Resolution Spherical Images
Authors: 
Wei Yu, Daren Zha, Nan Mu, Tianshu FuAuthors Info & Claims
Wei Yu, Daren Zha, Nan Mu, Tianshu FuAuthors Info & ClaimsPages 38 - 42
Abstract
With the prevalence of virtual reality, augmented reality and autonomous robots, the high resolution spherical images they produced make the standard convolutional neural networks (CNNs), which have been proven powerful on perspective images, non-trivial.
The classic solution to utilize CNNs on spherical images is to project the spherical images onto plane and learning the planar images using conventional CNNs. But the distortion generated by the projection of spherical images to planar images invalidates the projection based models. Besides, these models are not robust to rotations which are the basic transformation of spherical images. Another type of solution based on spherical harmonics recently proposed by Cohen et al. [1] is rotation equivariant, but can't handle high resolution spherical images with its expensive computational cost.
To process high resolution spherical images, we proposed the Hybrid CNNs. Our framework is both computationally efficient and rotation equivariant with two kinds of convolution operations defined in this paper. We compared our method with several baseline models in two classification tasks. The experimental results demonstrate the computational efficiency and rotation equivariance of the Hybrid CNNs.
References
[1]
Cohen, T. S., Geiger, M., Kohler, J., and Welling, M. 2018. Spherical CNNs. International Conference on Learning Representations.
[2]
Su, Y., and Grauman, K. 2017. Flat2Sphere: Learning Spherical Convolution for Fast Features from 360° Imagery. arXiv: Computer Vision and Pattern Recognition.
[3]
Ronchi, C., Iacono, R., and Paolucci, P. S. 1996. The "Cubed Sphere". Journal of Computational Physics, 124(1), 93--114.
[4]
Kondor, R., Lin, Z., and Trivedi, S. (2018). Clebsch-Gordan Nets: a Fully Fourier Space Spherical Convolutional Neural Network. 32nd Conference on Neural Information Processing Systems.
[5]
Driscoll, J. R., and Healy, D. M. 1994. Computing fourier transforms and convolutions on the 2-sphere. Advancesin Applied Mathematics.
[6]
Muciaccia, P. F., Natoli, P., and Vittorio, N. 1997. Fast Spherical Harmonic Analysis: A Quick Algorithm for Generating and/or InvertingFull-Sky, High-Resolution Cosmic Microwave Background Anisotropy Maps. The Astrophysical Journal, 488(2).
[7]
Saff, E. B., and Kuijlaars, A. B. 1997. Distributing many points on a sphere. The Mathematical Intelligencer, 19(1), 5--11.
[8]
Gorski, K.M., E. Hivon, A.J. Banday, B.D. Wandelt, F.K. Hansen, M. Reinecke, and M. Bartelmann. 2005. HEALPix: A Framework for High-resolution Discretization and Fast Analysis of Data Distributed on the Sphere, Ap.J., 622, 759--771.
[9]
Gorski, K.M., E. Hivon, A.J. Banday, B.D. Wandelt, F.K. Hansen, M. Reinecke, and M. Bartelmann. http://healpix.sf.net
[10]
Lecun, Y., Bottou, L., Bengio, Y., and Haffner, P. 1998. Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86(11), 2278--2324.
[11]
Chang, A. X., Funkhouser, T. A., Guibas, L. J., Hanrahan, P., Huang, Q., Li, Z., ... and Yu, F. 2015. ShapeNet: An Information-Rich 3D Model Repository. arXiv: Graphics.
[12]
Coors, B., Condurache, A. P., and Geiger, A. (2018). SphereNet: Learning Spherical Representations for Detection and Classification in Omnidirectional Images. european conference on computer vision.
Index Terms
- Hybrid CNNs: A Rotation Equivariant Framework for High Resolution Spherical Images
Recommendations
Convolutional neural networks for wavelet domain super resolution
Proposed a super resolution method with higher reconstruction accuracy than before.Cast super resolution as a problem of estimating sparse wavelet detail coefficients.Estimated sparse wavelet coefficients using a convolutional neural network (CNN)...
Deep learning for multisensor image resolution enhancement
GeoAI '17: Proceedings of the 1st Workshop on Artificial Intelligence and Deep Learning for Geographic Knowledge DiscoveryWe describe a deep learning convolutional neural network (CNN) for enhancing low resolution multispectral satellite imagery without the use of a panchromatic image. For training, low resolution images are used as input and corresponding high resolution ...
Group emotion recognition with individual facial emotion CNNs and global image based CNNs
ICMI '17: Proceedings of the 19th ACM International Conference on Multimodal InteractionThis paper presents our approach for group-level emotion recognition in the Emotion Recognition in the Wild Challenge 2017. The task is to classify an image into one of the group emotion such as positive, neutral or negative. Our approach is based on ...
Comments
Information & Contributors
Information
Published In
February 2019
170 pages
ISBN:9781450362047
DOI:10.1145/3316551
Copyright © 2019 ACM.
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from [email protected]
Publisher
Association for Computing Machinery
New York, NY, United States
Publication History
Published: 24 February 2019
Check for updates
Author Tags
Qualifiers
- Research-article
- Research
- Refereed limited
Conference
ICDSP 2019
ICDSP 2019: 2019 3rd International Conference on Digital Signal Processing
February 24 - 26, 2019
Jeju Island, Republic of Korea
Contributors
Other Metrics
Bibliometrics & Citations
Bibliometrics
Article Metrics
- 0Total Citations
- 83Total Downloads
- Downloads (Last 12 months)0
- Downloads (Last 6 weeks)0
Reflects downloads up to 05 Mar 2025
Other Metrics
Citations
View Options
Login options
Check if you have access through your login credentials or your institution to get full access on this article.
Sign in