mlsanityData to 14 Oct 2026

EMCStereo: Attention-Enhanced Stereo Matching for Thin-Structure Depth Estimation with a Synthetic Tree-Branch Benchmark

arXiv 2609.13233 ↗First posted 2 Sep 2026cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Thin structures such as tree branches are among the hardest cases for stereo matching: a branch is only a few pixels wide, the background is cluttered, and dense ground truth for real branches is nearly impossible to label by hand. We make three contributions. First, EMCStereo integrates three lightweight attention modules into a PSMNet-style cost-volume backbone: Efficient Multi-scale Attention (EMA) on deep semantic features, a Multi-Scale Fusion block (MSFblock) learning spatial pyramid weights instead of concatenating them, and Coordinate Attention (CoordAtt) on final matching features. Because MSFblock collapses four pyramid branches into one, the modules leave the network 2.0% smaller and add only 1.7% inference time overhead. Second, VirtualTree is a synthetic stereo dataset rendered in Unreal Engine 5 with a simulated ZED Mini rig, providing 5,520 pairs with exact disparity for thin branches. Third, an eight-way ablation establishes a run-to-run noise floor of 0.009 px end-point error (EPE). EMCStereo achieves 1.31 px EPE (5.96% D1-all) on the VirtualTree test split, 1.00 px on SceneFlow, and 0.80, 0.73, 0.62, and 3.19 px on KITTI 2012, KITTI 2015, ETH3D, and Middlebury, with depth accuracy delta_1 from 92.6% to 98.7%. Evaluated against the noise floor, the attention stack is accuracy-neutral at a 100-epoch budget, while MSFblock and CoordAtt cost 0.03-0.05 px unless EMA is present.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 22 of its 31 references that are in the map. First posted on arXiv 2 Sep 2026.

Builds on

22 of 31 references.

References · 22

  1. 2017Squeeze-and-Excitation Networkscited by 4,789
  2. 2018CBAM: Convolutional Block Attention Modulecited by 2,477
  3. 2015A Large Dataset to Train Convolutional Networks for Disparity, Optical Flow, and Scene Flow Estimationcited by 881
  4. 2018Pyramid Stereo Matching Networkcited by 489
  5. 2017End-to-End Learning of Geometry and Context for Deep Stereo Regressioncited by 360
  6. 2021RAFT-Stereo: Multilevel Recurrent Field Transforms for Stereo Matchingcited by 353
  7. 2019GA-Net: Guided Aggregation Net for End-to-end Stereo Matchingcited by 244
  8. 2021Coordinate Attention for Efficient Mobile Network Designcited by 241
  9. 2019Group-wise Correlation Stereo Networkcited by 235
  10. 2020AANet: Adaptive Aggregation Network for Efficient Stereo Matchingcited by 202
  11. 2015Stereo Matching by Training a Convolutional Neural Network to Compare Image Patchescited by 151
  12. 2022Attention Concatenation Volume for Accurate and Efficient Stereo Matchingcited by 130
10 more
  1. 2019DeepPruner: Learning Efficient Stereo Matching via Differentiable PatchMatchcited by 114
  2. 2018SegStereo: Exploiting Semantic Information for Disparity Estimationcited by 88
  3. 2017Cascade Residual Learning: A Two-stage Convolutional Neural Network for Stereo Matchingcited by 63
  4. 2025DEFOM-Stereo: Depth Foundation Model Based Stereo Matchingcited by 52
  5. 2023Efficient Multi-Scale Attention Module with Cross-Spatial Learningcited by 41
  6. 2018Practical Deep Stereo (PDS): Toward applications-friendly deep stereo matchingcited by 20
  7. 2023SHISRCNet: Super-resolution And Classification Network For Low-resolution Breast Cancer Histopathology Imagecited by 7
  8. 2025Performance Evaluation of Deep Learning for Tree Branch Segmentation in Autonomous Forestry Systemscited by 7
  9. 2025YOLO and SGBM Integration for Autonomous Tree Branch Detection and Depth Estimation in Radiata Pine Pruning Applicationscited by 6
  10. 2025Genetic Algorithms For Parameter Optimization for Disparity Map Generation of Radiata Pine Branch Imagescited by 4

Built on by

Not cited yet.