PBCStereo: A Compressed Stereo Network with Pure Binary Convolutional Operations

Jiaxuan Cai, ZHI QI, Keqi Fu, Xulong Shi, Zan Li, Xuanyu Liu, Hao Liu; Proceedings of the Asian Conference on Computer Vision (ACCV), 2022, pp. 4378-4394

Abstract

Although end-to-end stereo matching networks achieve great performance for disparity estimation, most of them require far too many floating-point operations to deploying on resource-constrained devices. To solve this problem, we propose PBCStereo, the first lightweight stereo network using pure binarized convolutional operations. The degradation of feature diversity, which is aggravated by binary deconvolution, is alleviated via our novel upsampling module (IBC). Furthermore, we propose an effective coding method, named BIL, for the insufficient binarization of the input layer. Based on IBC modules and BIL coding, all convolutional operations become binary in our stereo matching pipeline. PBCStereo gets 39x saving in OPs while achieving comparable accuracy on SceneFlow and KITTI datasets.

Related Material

[pdf] [code]
[bibtex]
@InProceedings{Cai_2022_ACCV, author = {Cai, Jiaxuan and QI, ZHI and Fu, Keqi and Shi, Xulong and Li, Zan and Liu, Xuanyu and Liu, Hao}, title = {PBCStereo: A Compressed Stereo Network with Pure Binary Convolutional Operations}, booktitle = {Proceedings of the Asian Conference on Computer Vision (ACCV)}, month = {December}, year = {2022}, pages = {4378-4394} }