Method

Variational Iterative Stereo: Posterior-Guided Lookup with Calibrated Uncertainty [vis-stereo]


Submitted on 13 Jun. 2026 05:14 by
Junjie Zhang (China University of Mining and Technology)

Running time:0.48 s
Environment:GPU @ 2.5 Ghz (Python)

Method Description:
VIS-Stereo is an iterative stereo matching method
built on the IGEV++ backbone. At each refinement
step it keeps a Gaussian posterior over disparity
instead of a single estimate, and reads the cost
volume through a window whose width is set by the
predicted variance, so the lookup gathers more
context in ambiguous regions and stays sharp
elsewhere. The variance is propagated across
iterations and tightens as evidence accumulates,
with a closed-form KL term keeping its trajectory
stable. It is supervised against the running
disparity error, so the same variance also
provides a per-pixel uncertainty that tracks the
actual error. The feature extractor, cost volume,
and GRU update block are unchanged from IGEV++,
and the added heads increase the parameter count
by 0.6%; inference runs in PyTorch on a single
NVIDIA RTX 4090.
Parameters:
16 GRU update iterations at inference; cost-volume
lookup radius R = 4. Loss weights: variance term
λ_σ = 0.1, KL continuity term β_KL = 0.05.
Trained with AdamW and a one-cycle schedule (peak
learning rate 2×10⁻⁴). Backbone settings follow
IGEV++.
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 2.48 1.52
All / Est 1.33 2.48 1.52
Noc / All 1.22 2.45 1.42
Noc / Est 1.22 2.45 1.42
This table as LaTeX

Test Image 0

Error D1-bg D1-fg D1-all
All / All 1.64 0.99 1.55
All / Est 1.64 0.99 1.55
Noc / All 1.59 0.99 1.51
Noc / Est 1.59 0.99 1.51
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 1

Error D1-bg D1-fg D1-all
All / All 1.59 2.90 1.74
All / Est 1.59 2.90 1.74
Noc / All 1.52 2.90 1.68
Noc / Est 1.52 2.90 1.68
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 2

Error D1-bg D1-fg D1-all
All / All 1.90 7.22 2.16
All / Est 1.90 7.22 2.16
Noc / All 1.84 7.22 2.10
Noc / Est 1.84 7.22 2.10
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 3

Error D1-bg D1-fg D1-all
All / All 2.13 1.90 2.11
All / Est 2.13 1.90 2.11
Noc / All 2.13 1.90 2.11
Noc / Est 2.13 1.90 2.11
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 4

Error D1-bg D1-fg D1-all
All / All 0.55 0.36 0.52
All / Est 0.55 0.36 0.52
Noc / All 0.55 0.36 0.52
Noc / Est 0.55 0.36 0.52
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 5

Error D1-bg D1-fg D1-all
All / All 1.97 2.83 2.05
All / Est 1.97 2.83 2.05
Noc / All 1.88 2.83 1.96
Noc / Est 1.88 2.83 1.96
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 6

Error D1-bg D1-fg D1-all
All / All 1.89 0.96 1.79
All / Est 1.89 0.96 1.79
Noc / All 1.93 0.96 1.82
Noc / Est 1.93 0.96 1.82
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 7

Error D1-bg D1-fg D1-all
All / All 0.25 3.75 0.93
All / Est 0.25 3.75 0.93
Noc / All 0.25 3.75 0.94
Noc / Est 0.25 3.75 0.94
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 8

Error D1-bg D1-fg D1-all
All / All 0.22 2.37 0.62
All / Est 0.22 2.37 0.62
Noc / All 0.22 2.37 0.62
Noc / Est 0.22 2.37 0.62
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 9

Error D1-bg D1-fg D1-all
All / All 0.28 1.38 0.56
All / Est 0.28 1.38 0.56
Noc / All 0.28 1.45 0.57
Noc / Est 0.28 1.45 0.57
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 10

Error D1-bg D1-fg D1-all
All / All 1.03 2.51 1.37
All / Est 1.03 2.51 1.37
Noc / All 1.04 2.51 1.38
Noc / Est 1.04 2.51 1.38
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 11

Error D1-bg D1-fg D1-all
All / All 0.71 0.61 0.69
All / Est 0.71 0.61 0.69
Noc / All 0.71 0.61 0.69
Noc / Est 0.71 0.61 0.69
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 12

Error D1-bg D1-fg D1-all
All / All 0.68 0.78 0.68
All / Est 0.68 0.78 0.68
Noc / All 0.53 0.78 0.55
Noc / Est 0.53 0.78 0.55
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 13

Error D1-bg D1-fg D1-all
All / All 0.52 0.11 0.47
All / Est 0.52 0.11 0.47
Noc / All 0.48 0.11 0.43
Noc / Est 0.48 0.11 0.43
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 14

Error D1-bg D1-fg D1-all
All / All 1.38 0.00 1.36
All / Est 1.38 0.00 1.36
Noc / All 1.21 0.00 1.18
Noc / Est 1.21 0.00 1.18
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 15

Error D1-bg D1-fg D1-all
All / All 2.51 0.26 2.30
All / Est 2.51 0.26 2.30
Noc / All 2.56 0.26 2.35
Noc / Est 2.56 0.26 2.35
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 16

Error D1-bg D1-fg D1-all
All / All 3.30 0.24 2.85
All / Est 3.30 0.24 2.85
Noc / All 3.09 0.24 2.67
Noc / Est 3.09 0.24 2.67
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 17

Error D1-bg D1-fg D1-all
All / All 1.07 0.20 0.98
All / Est 1.07 0.20 0.98
Noc / All 1.07 0.20 0.97
Noc / Est 1.07 0.20 0.97
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 18

Error D1-bg D1-fg D1-all
All / All 4.45 1.92 3.25
All / Est 4.45 1.92 3.25
Noc / All 4.39 1.92 3.21
Noc / Est 4.39 1.92 3.21
This table as LaTeX

Input Image

D1 Result

D1 Error


Test Image 19

Error D1-bg D1-fg D1-all
All / All 0.66 0.96 0.69
All / Est 0.66 0.96 0.69
Noc / All 0.67 0.96 0.70
Noc / Est 0.67 0.96 0.70
This table as LaTeX

Input Image

D1 Result

D1 Error




eXTReMe Tracker