Unsupervised Optical Flow Estimation Based on Improved Feature Pyramid
- PDF / 1,376,939 Bytes
- 12 Pages / 439.37 x 666.142 pts Page_size
- 83 Downloads / 235 Views
Unsupervised Optical Flow Estimation Based on Improved Feature Pyramid Bo Yang1 · Huan Xie1 · Hongbin Li2 · Nuohan Li3 · Anchang Liu2 · Zhigang Ren1 · Kuan Ye1 · Rong Zhu1 · Xuezhi Xiang4
© Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Deep learning methods for optical flow estimation usually increase the receptive field of convolution through reducing image resolution, which results in loss of spatial detail information during feature extraction. In this paper, we introduce dilated convolution into feature pyramid network, which can extract multi-scale features containing more motion details and can further improve the accuracy of optical flow estimation. The unsupervised loss function is based on forward–backward consistency check and robust census transform that has a good constraint performance in the case of illumination changes, which can train an unsupervised learning optical flow model with higher accuracy. Our network is trained on FlyingChairs and KITTI raw datasets with an unsupervised manner and tested on MPI-Sintel, KITTI 2012 and KITTI 2015 benchmarks. The experimental results show the advantages of our method in unsupervised learning approaches. Keywords Optical flow estimation · Deep learning · Feature pyramid · Dilated convolution
1 Introduction Optical flow estimation is an important technical means for processing dynamic image sequences, which has a range of applications. Horn and Schunck [1] first use variational method to estimate optical flow, which lay a foundation for the work of traditional optical
This work was supported in part by the State Grid Corporation Science and Technology Foundation under Grant 52022318001N, in part by the National Natural Science Foundation of China under Grant 61401113, and in part by the Natural Science Foundation of Heilongjiang Province of China under Grant LC201426.
B
Xuezhi Xiang [email protected]
1
The Electric Power Research Institute of State Grid Beijing Electric Power Company, Beijing, China
2
State Grid Beijing Electric Power Company, Beijing, China
3
Beijing Jindian United Power Supply Consulting Co., Ltd., Beijing, China
4
School of Information and Communication Engineering, Harbin Engineering University, Harbin 150001, China
123
B. Yang et al.
flow research. Most of the traditional methods improve the accuracy of optical flow estimation by adding a variety of new constraints to energy function, which can easily integrate the prior knowledge of motion. But this kind of methods is always computationally expensive and not suitable for real-time application. Recently, with the development of deep learning, convolutional neural network (CNN) has made great breakthroughs in many fields of computer vision, such as target recognition [2], image classification [3], semantic segmentation [4], and its influence has gradually seeped to optical flow estimation. Dosvitskiy et al. [5] first propose two end-to-end CNN models for optical flow estimation, which are called FlowNetS and FlowNetC, using the generic
Data Loading...
