Abstract
In the design of the traditional CNN model, there is always a balance between the spatial dimension and the number of channels. The high-dimensional spatial resolution is to preserve more detailed local information, while the large number of channels ensures more complex feature representation. For the current Light-field depth estimation algorithm, the designers usually choose to maintain a high spatial dimension in the network to improve the accuracy of the depth estimation, which results in a situation where the model has a large size and will take up huge amount of computing resources in the depth estimation process. In this paper, we introduce an effective pooling method: spectral pooling, to improve the problems of the original Light-field depth estimation network mentioned above. We transfer the light field feature map to the frequency domain through Fourier transform and reasonably reduce the size of the feature map in the frequency domain to an arbitrary size. The method can be used to reduce the complexity of the network and speed up training with similar performance to the original algorithm. More importantly, the new model with less memory demand can be better applied to light mobile device.
This work is partially supported by the Fundamental Research Funds for the Central Universities (2018RC54), and partially by the Fundamental Research Funds for the New Start Plan Project of Beijing Union University (Zk10201604).
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Rippel, O., Snoek, J., Adams, R.P.: Spectral representations for convolutional neural networks. In: Advances in Neural Information Processing Systems, pp. 2449–2457 (2015)
Khan, S.H., Hayat, M., Porikli, F.: Regularization of deep neural networks with spectral dropout. Neural Netw. 110, 82–90 (2019)
Wang, Z., Lan, Q., Huang, D., Wen, M.: Combining FFT and spectral-pooling for efficient convolution neural network model. In: 2016 2nd International Conference on Artificial Intelligence and Industrial Engineering (AIIE 2016). Atlantis Press (2016)
Gortler, S.J., Grzeszczuk, R., Szeliski, R., Cohen, M.F.: The lumigraph. In: Siggraph, vol. 96, pp. 43–54 (1996)
Levoy, M., Hanrahan, P.: Light field rendering. In: Proceedings of the 23rd Annual Conference on Computer Graphics and Interactive Techniques, pp. 31–42. ACM (1996)
Wanner, S., Goldluecke, B.: Variational light field analysis for disparity estimation and super-resolution. IEEE Trans. Pattern Anal. Mach. Intell. 36, 606–619 (2014)
Zhang, S., Sheng, H., Li, C., Zhang, J., Xiong, Z.: Robust depth estimation for light field via spinning parallelogram operator. Comput. Vis. Image Underst. 145, 148–159 (2016)
Shin, C., Jeon, H.-G., Yoon, Y., So Kweon, I., Joo Kim, S.: Epinet: a fully-convolutional neural network using epipolar geometry for depth from light field images. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 4748–4757 (2018)
Ng, R., Levoy, M., Brédif, M., Duval, G., Horowitz, M., Hanrahan, P., et al.: Light field photography with a hand-held plenoptic camera. Comput. Sci. Tech. Rep. CSTR 2, 1–11 (2005)
Wang, T.-C., Efros, A.A., Ramamoorthi, R.: Occlusion-aware depth estimation using light-field cameras. In: Proceedings of the IEEE International Conference on Computer Vision, pp. 3487–3495 (2015)
Heber, S., Yu, W., Pock, T.: Neural EPI-volume networks for shape from light field. In: Proceedings of the IEEE International Conference on Computer Vision, pp. 2252–2260 (2017)
Anisimov, Y., Didier, S.: Fast and efficient depth map estimation from light fields. In: 2017 International Conference on 3D Vision (3DV), pp. 337–346. IEEE (2017)
Cooley, J.W., Tukey, J.W.: An algorithm for the machine calculation of complex Fourier series. Math. Comput. 19, 297–301 (1965)
Dansereau, D.G., Pizarro, O., Williams, S.B.: Linear volumetric focus for light field cameras. ACM Trans. Graph. 34, 15-1 (2015)
Dosovitskiy, A., et al.: Flownet: learning optical flow with convolutional networks. In: Proceedings of the IEEE International Conference on Computer Vision, pp. 2758–2766 (2015)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2019 Springer Nature Singapore Pte Ltd.
About this paper
Cite this paper
Zhang, X., Yang, G., Jiang, J. (2019). Spectral Pooling Based CNN Model Compression for Light Field Depth-Estimation. In: Wang, Y., Huang, Q., Peng, Y. (eds) Image and Graphics Technologies and Applications. IGTA 2019. Communications in Computer and Information Science, vol 1043. Springer, Singapore. https://doi.org/10.1007/978-981-13-9917-6_32
Download citation
DOI: https://doi.org/10.1007/978-981-13-9917-6_32
Published:
Publisher Name: Springer, Singapore
Print ISBN: 978-981-13-9916-9
Online ISBN: 978-981-13-9917-6
eBook Packages: Computer ScienceComputer Science (R0)