Method

Multi-frequency Adaptive Fusion Network for Real-time Stereo Matching [MAFNet++]


Submitted on 17 Sep. 2025 04:31 by
Ao Xu (Tongji University)

Running time:0.03 s
Environment:1 core @ 2.5 Ghz (C/C++)

Method Description:
Existing stereo matching networks typically rely
on either cost-volume construction based on 3D
convolutions or deformation methods based on
iterative optimization. The former incurs
significant computational overhead during cost
aggregation, whereas the latter often lacks the
ability to model non-local contextual information.
These methods exhibit poor compatibility on
resource-constrained mobile devices, limiting
their deployment in real-time applications. To
address this, we propose a Multi-frequency
Adaptive Fusion Network (MAFNet), which can
produce high-quality disparity maps using only
efficient 2D convolutions. Specifically, we design
an adaptive frequency-domain filtering attention
module that decomposes the full cost volume into
high-frequency and low-frequency volumes,
performing frequency-aware feature aggregation
separately.
Parameters:
The proposed MAFNet model was implemented using
PyTorch, and experiments were carried out on the
Scene Flow, KITTI 2012, and KITTI 2015 datasets.
All experiments were conducted on an NVIDIA RTX
5090 GPU. We employed the AdamW [21] optimizer
combined with a one-cycle learning rate scheduler
with warm-up, setting the maximum learning rate to
8e-4. The weights in the loss function were
configured as \lambda_0\ =0.3 and \lambda_1\ =1.0.
During training, we initially pre-trained on the
Scene Flow dataset for 200k steps with a batch
size of 16 to adequately learn the distribution of
large-scale synthetic data. The resulting model
was then fine-tuned for 50k steps on the mixed
KITTI 2012 and KITTI 2015 training sets to improve
performance in real-world environments. During
training, each input image was randomly cropped to
a resolution of 256×512.
Latex Bibtex:

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 1.33 2.53 1.53
All / Est 1.33 2.53 1.53
Noc / All 1.21 2.52 1.43
Noc / Est 1.21 2.52 1.43
This table as LaTeX

Test Image 0

Error D1-bg D1-fg D1-all
All / All 1.30 1.77 1.37
All / Est 1.30 1.77 1.37
Noc / All 1.30 1.77 1.36
Noc / Est 1.30 1.77 1.36
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 1

Error D1-bg D1-fg D1-all
All / All 1.61 1.87 1.64
All / Est 1.61 1.87 1.64
Noc / All 1.54 1.87 1.58
Noc / Est 1.54 1.87 1.58
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 2

Error D1-bg D1-fg D1-all
All / All 1.93 7.18 2.19
All / Est 1.93 7.18 2.19
Noc / All 1.88 7.18 2.14
Noc / Est 1.88 7.18 2.14
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 3

Error D1-bg D1-fg D1-all
All / All 1.71 0.44 1.59
All / Est 1.71 0.44 1.59
Noc / All 1.69 0.44 1.58
Noc / Est 1.69 0.44 1.58
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 4

Error D1-bg D1-fg D1-all
All / All 0.34 0.20 0.32
All / Est 0.34 0.20 0.32
Noc / All 0.34 0.20 0.31
Noc / Est 0.34 0.20 0.31
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 5

Error D1-bg D1-fg D1-all
All / All 1.91 2.86 2.00
All / Est 1.91 2.86 2.00
Noc / All 1.85 2.86 1.94
Noc / Est 1.85 2.86 1.94
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 6

Error D1-bg D1-fg D1-all
All / All 2.45 1.74 2.38
All / Est 2.45 1.74 2.38
Noc / All 2.50 1.74 2.42
Noc / Est 2.50 1.74 2.42
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 7

Error D1-bg D1-fg D1-all
All / All 0.26 2.94 0.79
All / Est 0.26 2.94 0.79
Noc / All 0.26 2.94 0.80
Noc / Est 0.26 2.94 0.80
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 8

Error D1-bg D1-fg D1-all
All / All 0.28 2.17 0.63
All / Est 0.28 2.17 0.63
Noc / All 0.28 2.17 0.63
Noc / Est 0.28 2.17 0.63
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 9

Error D1-bg D1-fg D1-all
All / All 0.33 0.92 0.48
All / Est 0.33 0.92 0.48
Noc / All 0.33 0.96 0.49
Noc / Est 0.33 0.96 0.49
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 10

Error D1-bg D1-fg D1-all
All / All 1.22 2.56 1.52
All / Est 1.22 2.56 1.52
Noc / All 1.23 2.56 1.54
Noc / Est 1.23 2.56 1.54
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 11

Error D1-bg D1-fg D1-all
All / All 0.89 0.87 0.89
All / Est 0.89 0.87 0.89
Noc / All 0.89 0.87 0.89
Noc / Est 0.89 0.87 0.89
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 12

Error D1-bg D1-fg D1-all
All / All 0.61 1.30 0.66
All / Est 0.61 1.30 0.66
Noc / All 0.47 1.30 0.52
Noc / Est 0.47 1.30 0.52
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 13

Error D1-bg D1-fg D1-all
All / All 0.46 0.29 0.44
All / Est 0.46 0.29 0.44
Noc / All 0.45 0.29 0.43
Noc / Est 0.45 0.29 0.43
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 14

Error D1-bg D1-fg D1-all
All / All 1.44 0.00 1.41
All / Est 1.44 0.00 1.41
Noc / All 1.29 0.00 1.27
Noc / Est 1.29 0.00 1.27
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 15

Error D1-bg D1-fg D1-all
All / All 2.24 0.42 2.07
All / Est 2.24 0.42 2.07
Noc / All 2.28 0.42 2.11
Noc / Est 2.28 0.42 2.11
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 16

Error D1-bg D1-fg D1-all
All / All 3.29 0.03 2.81
All / Est 3.29 0.03 2.81
Noc / All 3.08 0.03 2.63
Noc / Est 3.08 0.03 2.63
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 17

Error D1-bg D1-fg D1-all
All / All 0.87 0.18 0.80
All / Est 0.87 0.18 0.80
Noc / All 0.85 0.18 0.78
Noc / Est 0.85 0.18 0.78
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 18

Error D1-bg D1-fg D1-all
All / All 4.43 1.77 3.17
All / Est 4.43 1.77 3.17
Noc / All 4.36 1.77 3.12
Noc / Est 4.36 1.77 3.12
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 19

Error D1-bg D1-fg D1-all
All / All 0.67 0.69 0.67
All / Est 0.67 0.69 0.67
Noc / All 0.67 0.69 0.67
Noc / Est 0.67 0.69 0.67
This table as LaTeX

Input Image

D1 Result

D1 Error




eXTReMe Tracker