Method

Hadamard Attention Recurrent Transformer: A Strong Baseline for Stereo Matching Transformer [HART]
https://github.com/ZYangChen/HART

Submitted on 1 Feb. 2024 04:53 by
Ziyang Chen (Guizhou University)

Running time:0.25 s
Environment:NVIDIA Tesla A100 (PyTorch)

Method Description:
In light of the advancements in transformer
technology, existing research posits the
construction of stereo transformers as a potential
solution to the binocular stereo matching
challenge. However, constrained by the low-rank
bottleneck and quadratic complexity of attention
mechanisms, stereo transformers still fail to
demonstrate sufficient nonlinear expressiveness
within a reasonable inference time. The lack of
focus on key homonymous points renders the
representations of such methods vulnerable to
challenging conditions, including reflections and
weak textures. Furthermore, a slow computing speed
is not conducive to the application. To overcome
these difficulties, we present the Hadamard
Attention Recurrent Stereo Transformer (HART).
Parameters:
use mixed precision & True
batch size used during training & 8
crop size & 320 $\times$ 720
max learning rate & 0.0002
length of training schedule & 200000
recurrent-number during training & 22
Weight decay in optimizer & 0.00001
Latex Bibtex:
@article{chen2025hart,
title={Hadamard Attention Recurrent Transformer:
A Strong Baseline for Stereo Matching Transformer},
author={Chen, Ziyang and Zhang, Yongjun and Li,
Wenting and Wang, Bingshu and Wu, Yabo and Zhao,
Yong and Chen, CL},
journal={arXiv preprint arXiv:2501.01023},
year={2025}
}

Detailed Results

This page provides detailed results for the method(s) selected. For the first 20 test images, the percentage of erroneous pixels is depicted in the table. We use the error metric described in Object Scene Flow for Autonomous Vehicles (CVPR 2015), which considers a pixel to be correctly estimated if the disparity or flow end-point error is <3px or <5% (for scene flow this criterion needs to be fulfilled for both disparity maps and the flow map). Underneath, the left input image, the estimated results and the error maps are shown (for disp_0/disp_1/flow/scene_flow, respectively). The error map uses the log-color scale described in Object Scene Flow for Autonomous Vehicles (CVPR 2015), depicting correct estimates (<3px or <5% error) in blue and wrong estimates in red color tones. Dark regions in the error images denote the occluded pixels which fall outside the image boundaries. The false color maps of the results are scaled to the largest ground truth disparity values / flow magnitudes.

Test Set Average

Error D1-bg D1-fg D1-all
All / All 1.39 2.49 1.57
All / Est 1.39 2.49 1.57
Noc / All 1.29 2.50 1.49
Noc / Est 1.29 2.50 1.49
This table as LaTeX

Test Image 0

Error D1-bg D1-fg D1-all
All / All 1.67 0.34 1.49
All / Est 1.67 0.34 1.49
Noc / All 1.66 0.34 1.48
Noc / Est 1.66 0.34 1.48
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 1

Error D1-bg D1-fg D1-all
All / All 1.79 6.10 2.27
All / Est 1.79 6.10 2.27
Noc / All 1.72 6.10 2.22
Noc / Est 1.72 6.10 2.22
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 2

Error D1-bg D1-fg D1-all
All / All 2.69 9.71 3.03
All / Est 2.69 9.71 3.03
Noc / All 2.65 9.71 3.00
Noc / Est 2.65 9.71 3.00
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 3

Error D1-bg D1-fg D1-all
All / All 2.01 1.23 1.94
All / Est 2.01 1.23 1.94
Noc / All 2.01 1.23 1.94
Noc / Est 2.01 1.23 1.94
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 4

Error D1-bg D1-fg D1-all
All / All 0.75 0.21 0.66
All / Est 0.75 0.21 0.66
Noc / All 0.75 0.21 0.66
Noc / Est 0.75 0.21 0.66
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 5

Error D1-bg D1-fg D1-all
All / All 1.80 1.92 1.81
All / Est 1.80 1.92 1.81
Noc / All 1.66 1.92 1.69
Noc / Est 1.66 1.92 1.69
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 6

Error D1-bg D1-fg D1-all
All / All 2.66 1.63 2.55
All / Est 2.66 1.63 2.55
Noc / All 2.72 1.63 2.60
Noc / Est 2.72 1.63 2.60
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 7

Error D1-bg D1-fg D1-all
All / All 0.25 3.26 0.84
All / Est 0.25 3.26 0.84
Noc / All 0.26 3.26 0.85
Noc / Est 0.26 3.26 0.85
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 8

Error D1-bg D1-fg D1-all
All / All 0.26 2.54 0.68
All / Est 0.26 2.54 0.68
Noc / All 0.25 2.54 0.68
Noc / Est 0.25 2.54 0.68
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 9

Error D1-bg D1-fg D1-all
All / All 0.30 1.38 0.57
All / Est 0.30 1.38 0.57
Noc / All 0.30 1.45 0.58
Noc / Est 0.30 1.45 0.58
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 10

Error D1-bg D1-fg D1-all
All / All 1.13 1.88 1.30
All / Est 1.13 1.88 1.30
Noc / All 1.14 1.88 1.31
Noc / Est 1.14 1.88 1.31
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 11

Error D1-bg D1-fg D1-all
All / All 1.10 0.49 0.99
All / Est 1.10 0.49 0.99
Noc / All 1.11 0.49 1.00
Noc / Est 1.11 0.49 1.00
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 12

Error D1-bg D1-fg D1-all
All / All 0.69 0.96 0.70
All / Est 0.69 0.96 0.70
Noc / All 0.53 0.96 0.56
Noc / Est 0.53 0.96 0.56
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 13

Error D1-bg D1-fg D1-all
All / All 0.52 0.04 0.47
All / Est 0.52 0.04 0.47
Noc / All 0.51 0.04 0.45
Noc / Est 0.51 0.04 0.45
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 14

Error D1-bg D1-fg D1-all
All / All 1.52 0.11 1.50
All / Est 1.52 0.11 1.50
Noc / All 1.27 0.11 1.25
Noc / Est 1.27 0.11 1.25
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 15

Error D1-bg D1-fg D1-all
All / All 2.64 0.23 2.42
All / Est 2.64 0.23 2.42
Noc / All 2.70 0.23 2.47
Noc / Est 2.70 0.23 2.47
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 16

Error D1-bg D1-fg D1-all
All / All 3.37 0.08 2.88
All / Est 3.37 0.08 2.88
Noc / All 3.16 0.08 2.70
Noc / Est 3.16 0.08 2.70
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 17

Error D1-bg D1-fg D1-all
All / All 1.07 0.12 0.97
All / Est 1.07 0.12 0.97
Noc / All 1.06 0.12 0.96
Noc / Est 1.06 0.12 0.96
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 18

Error D1-bg D1-fg D1-all
All / All 5.16 1.15 3.26
All / Est 5.16 1.15 3.26
Noc / All 5.09 1.15 3.20
Noc / Est 5.09 1.15 3.20
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 19

Error D1-bg D1-fg D1-all
All / All 0.89 0.77 0.88
All / Est 0.89 0.77 0.88
Noc / All 0.89 0.77 0.88
Noc / Est 0.89 0.77 0.88
This table as LaTeX

Input Image

D1 Result

D1 Error




eXTReMe Tracker