# Separable Convolutions for Optimizing 3D Stereo Networks

arXiv: https://arxiv.org/abs/2108.10216 · HTML: https://mlsanity.com/paper/2108.10216/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Builds on 16 of its 25 references that are in the map.
- First posted on arXiv 23 Aug 2021.

## Builds on (most cited first, 16 of 16)

- [MobileNetV2: Inverted Residuals and Linear Bottlenecks](https://mlsanity.com/paper/1801.04381/) (2018)
- [ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices](https://mlsanity.com/paper/1707.01083/) (2017)
- [A Large Dataset to Train Convolutional Networks for Disparity, Optical Flow, and Scene Flow Estimation](https://mlsanity.com/paper/1512.02134/) (2015)
- [Pyramid Stereo Matching Network](https://mlsanity.com/paper/1803.08669/) (2018)
- [Learning Spatio-Temporal Representation with Pseudo-3D Residual Networks](https://mlsanity.com/paper/1711.10305/) (2017)
- [End-to-End Learning of Geometry and Context for Deep Stereo Regression](https://mlsanity.com/paper/1703.04309/) (2017)
- [GA-Net: Guided Aggregation Net for End-to-end Stereo Matching](https://mlsanity.com/paper/1904.06587/) (2019)
- [Stereo Matching by Training a Convolutional Neural Network to Compare Image Patches](https://mlsanity.com/paper/1510.05970/) (2015)
- [DeepPruner: Learning Efficient Stereo Matching via Differentiable PatchMatch](https://mlsanity.com/paper/1909.05845/) (2019)
- [SegStereo: Exploiting Semantic Information for Disparity Estimation](https://mlsanity.com/paper/1807.11699/) (2018)
- [Real-time self-adaptive deep stereo](https://mlsanity.com/paper/1810.05424/) (2018)
- [Anytime Stereo Image Depth Estimation on Mobile Devices](https://mlsanity.com/paper/1810.11408/) (2018)
- [EdgeStereo: A Context Integrated Residual Pyramid Network for Stereo Matching](https://mlsanity.com/paper/1803.05196/) (2018)
- [Practical Deep Stereo (PDS): Toward applications-friendly deep stereo matching](https://mlsanity.com/paper/1806.01677/) (2018)
- [3D Depthwise Convolution: Reducing Model Parameters in 3D Vision Tasks](https://mlsanity.com/paper/1808.01556/) (2018)
- [On the Synergies between Machine Learning and Binocular Stereo for Depth Estimation from Images: a Survey](https://mlsanity.com/paper/2004.08566/) (2020)

## Built on by (most cited first, 5 of 5)

- [Fast-FoundationStereo: Real-Time Zero-Shot Stereo Matching](https://mlsanity.com/paper/2512.11130/) (2025)
- [Bayesian Learning for Disparity Map Refinement for Semi-Dense Active Stereo Vision](https://mlsanity.com/paper/2209.05082/) (2022)
- [FCDSN-DC: An Accurate and Lightweight Convolutional Neural Network for Stereo Estimation with Depth Completion](https://mlsanity.com/paper/2209.06525/) (2022)
- [Distilling Stereo Networks for Performant and Efficient Leaner Networks](https://mlsanity.com/paper/2503.18544/) (2025)
- [LeanStereo: A Leaner Backbone based Stereo Network](https://mlsanity.com/paper/2503.18557/) (2025)

## Abstract (arXiv)

Deep learning based 3D stereo networks give superior performance compared to 2D networks and conventional stereo methods. However, this improvement in the performance comes at the cost of increased computational complexity, thus making these networks non-practical for the real-world applications. Specifically, these networks use 3D convolutions as a major work horse to refine and regress disparities. In this work first, we show that these 3D convolutions in stereo networks consume up to 94% of overall network operations and act as a major bottleneck. Next, we propose a set of "plug-&-run" separable convolutions to reduce the number of parameters and operations. When integrated with the existing state of the art stereo networks, these convolutions lead up to 7x reduction in number of operations and up to 3.5x reduction in parameters without compromising their performance. In fact these convolutions lead to improvement in their performance in the majority of cases.