# Fast Deep Stereo with 2D Convolutional Processing of Cost Signatures

arXiv: https://arxiv.org/abs/1903.04939 · HTML: https://mlsanity.com/paper/1903.04939/

## Key facts

- Cited by 12 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 8 Mar 2019.

## Built on by (most cited first, 12 of 12)

- [SimpleRecon: 3D Reconstruction Without 3D Convolutions](https://mlsanity.com/paper/2208.14743/) (2022)
- [MobileStereoNet: Towards Lightweight Deep Networks for Stereo Matching](https://mlsanity.com/paper/2108.09770/) (2021)
- [PVStereo: Pyramid Voting Module for End-to-End Self-Supervised Stereo Matching](https://mlsanity.com/paper/2103.07094/) (2021)
- [Fast-FoundationStereo: Real-Time Zero-Shot Stereo Matching](https://mlsanity.com/paper/2512.11130/) (2025)
- [A Survey on Deep Stereo Matching in the Twenties](https://mlsanity.com/paper/2407.07816/) (2024)
- [CVRecon: Rethinking 3D Geometric Feature Learning For Neural Reconstruction](https://mlsanity.com/paper/2304.14633/) (2023)
- [Towards Adversarially Robust and Domain Generalizable Stereo Matching by Rethinking DNN Feature Backbones](https://mlsanity.com/paper/2108.00335/) (2021)
- [MAFNet:Multi-frequency Adaptive Fusion Network for Real-time Stereo Matching](https://mlsanity.com/paper/2512.04358/) (2025)
- [Bayesian Learning for Disparity Map Refinement for Semi-Dense Active Stereo Vision](https://mlsanity.com/paper/2209.05082/) (2022)
- [Multi-Scale Cost Volumes Cascade Network for Stereo Matching](https://mlsanity.com/paper/2102.01940/) (2021)
- [ES-Net: An Efficient Stereo Matching Network](https://mlsanity.com/paper/2103.03922/) (2021)
- [LeanStereo: A Leaner Backbone based Stereo Network](https://mlsanity.com/paper/2503.18557/) (2025)

## Abstract (arXiv)

Modern neural network-based algorithms are able to produce highly accurate depth estimates from stereo image pairs, nearly matching the reliability of measurements from more expensive depth sensors. However, this accuracy comes with a higher computational cost since these methods use network architectures designed to compute and process matching scores across all candidate matches at all locations, with floating point computations repeated across a match volume with dimensions corresponding to both space and disparity. This leads to longer running times to process each image pair, making them impractical for real-time use in robots and autonomous vehicles. We propose a new stereo algorithm that employs a significantly more efficient network architecture. Our method builds an initial match cost volume using traditional matching costs that are fast to compute, and trains a network to estimate disparity from this volume. Crucially, our network only employs per-pixel and two-dimensional convolution operations: to summarize the match information at each location as a low-dimensional feature vector, and to spatially process these `cost-signature' features to produce a dense disparity map. Experimental results on the KITTI benchmark show that our method delivers competitive accuracy at significantly higher speeds---running at 48 frames per second on a modern GPU.