Rotation Invariant Local Binary Convolution Neural Networks

Xin Zhang, Li Liu, Yuxiang Xie, Jie Chen, Lingda Wu, Matti Pietikainen; Proceedings of the IEEE International Conference on Computer Vision (ICCV), 2017, pp. 1210-1219

Abstract


Although CNNs are unprecedentedly powerful to learn effective representations, they are still parameter expensive and limited by the lack of ability to handle with the orientation transformation of the input data. To alleviate this problem, we propose a new differential module, Local Binary orientation Module(LBoM), which is a combination of Local Binary Convolution (LBC)[19] and Active Rotating Filters (ARFs)[38]. With LBoMs, a deep architecture named Rotation Invariant Local Binary Convolution Neural Networks(RI-LBCNNs) is constructed. RI-LBCNNs can be easy implemented and LBoM can be naturally inserted to popular models without any extra modification to the optimisation process. Meanwhile, The proposed RI-LBCNNs thus can be easily trained end to end. Extensive experiments show that the updating with the proposed LBoMs leads to significant reduction of learnable parameters and the reasonable performance on three benchmarks.

Related Material


[pdf]
[bibtex]
@InProceedings{Zhang_2017_ICCV,
author = {Zhang, Xin and Liu, Li and Xie, Yuxiang and Chen, Jie and Wu, Lingda and Pietikainen, Matti},
title = {Rotation Invariant Local Binary Convolution Neural Networks},
booktitle = {Proceedings of the IEEE International Conference on Computer Vision (ICCV) Workshops},
month = {Oct},
year = {2017}
}