Method

Learning Optical Flow with Adaptive Graph Reasoning [AGFlow]
https://github.com/megvii-research/AGFlow

Submitted on 25 Aug. 2021 11:33 by
Kunming Luo (Megvii)

Running time:0.2 s
Environment:8 cores @ 2.5 Ghz (Python)

Method Description:
We introduce a novel graph-based approach, namely adaptive graph reasoning for optical flow, i.e., AGFlow, which embeds graph techniques onto the matching pipeline to enable an effective context reasoning and information interaction. The proposed AGFlow learns to match features conditioned on scene context, and allows objects’ spatial extents to be well aggregated and thus largely decreases the uncertainty of ambiguous matching.
Parameters:
k=128
Latex Bibtex:
@InProceedings{luo2022learning,
title={Learning Optical Flow with Adaptive Graph Reasoning},
author={Luo, Ao and Yang, Fan and Luo, Kunming and Li, Xin and Fan, Haoqiang and Liu, Shuaicheng},
booktitle={AAAI},
year={2022},
}

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 Fl-bg Fl-fg Fl-all
All / All 4.52 6.75 4.89
All / Est 4.52 6.75 4.89
Noc / All 2.79 3.77 2.97
Noc / Est 2.79 3.77 2.97
This table as LaTeX

Test Image 0

Error Fl-bg Fl-fg Fl-all
All / All 1.72 16.79 3.79
All / Est 1.72 16.79 3.79
Noc / All 1.82 16.79 4.09
Noc / Est 1.82 16.79 4.09
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 1

Error Fl-bg Fl-fg Fl-all
All / All 1.56 6.18 2.07
All / Est 1.56 6.18 2.07
Noc / All 1.58 6.18 2.15
Noc / Est 1.58 6.18 2.15
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 2

Error Fl-bg Fl-fg Fl-all
All / All 5.56 10.59 5.81
All / Est 5.56 10.59 5.81
Noc / All 4.95 10.59 5.28
Noc / Est 4.95 10.59 5.28
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 3

Error Fl-bg Fl-fg Fl-all
All / All 6.93 41.99 10.17
All / Est 6.93 41.99 10.17
Noc / All 4.55 22.60 6.11
Noc / Est 4.55 22.60 6.11
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 4

Error Fl-bg Fl-fg Fl-all
All / All 2.31 2.27 2.30
All / Est 2.31 2.27 2.30
Noc / All 2.40 1.78 2.29
Noc / Est 2.40 1.78 2.29
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 5

Error Fl-bg Fl-fg Fl-all
All / All 4.14 0.10 3.78
All / Est 4.14 0.10 3.78
Noc / All 3.31 0.10 2.98
Noc / Est 3.31 0.10 2.98
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 6

Error Fl-bg Fl-fg Fl-all
All / All 4.90 0.53 4.44
All / Est 4.90 0.53 4.44
Noc / All 2.63 0.53 2.38
Noc / Est 2.63 0.53 2.38
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 7

Error Fl-bg Fl-fg Fl-all
All / All 1.24 3.75 1.73
All / Est 1.24 3.75 1.73
Noc / All 1.24 3.64 1.69
Noc / Est 1.24 3.64 1.69
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 8

Error Fl-bg Fl-fg Fl-all
All / All 0.27 2.44 0.67
All / Est 0.27 2.44 0.67
Noc / All 0.27 2.44 0.67
Noc / Est 0.27 2.44 0.67
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 9

Error Fl-bg Fl-fg Fl-all
All / All 0.45 2.46 0.96
All / Est 0.45 2.46 0.96
Noc / All 0.45 2.46 0.96
Noc / Est 0.45 2.46 0.96
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 10

Error Fl-bg Fl-fg Fl-all
All / All 1.19 3.19 1.65
All / Est 1.19 3.19 1.65
Noc / All 1.28 3.19 1.78
Noc / Est 1.28 3.19 1.78
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 11

Error Fl-bg Fl-fg Fl-all
All / All 2.65 0.82 2.32
All / Est 2.65 0.82 2.32
Noc / All 2.71 0.82 2.33
Noc / Est 2.71 0.82 2.33
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 12

Error Fl-bg Fl-fg Fl-all
All / All 1.43 0.74 1.38
All / Est 1.43 0.74 1.38
Noc / All 1.15 0.74 1.12
Noc / Est 1.15 0.74 1.12
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 13

Error Fl-bg Fl-fg Fl-all
All / All 3.40 1.65 3.19
All / Est 3.40 1.65 3.19
Noc / All 1.54 0.99 1.48
Noc / Est 1.54 0.99 1.48
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 14

Error Fl-bg Fl-fg Fl-all
All / All 2.79 1.07 2.76
All / Est 2.79 1.07 2.76
Noc / All 1.90 1.07 1.88
Noc / Est 1.90 1.07 1.88
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 15

Error Fl-bg Fl-fg Fl-all
All / All 10.11 3.08 9.47
All / Est 10.11 3.08 9.47
Noc / All 7.44 3.08 6.93
Noc / Est 7.44 3.08 6.93
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 16

Error Fl-bg Fl-fg Fl-all
All / All 13.92 1.93 12.16
All / Est 13.92 1.93 12.16
Noc / All 7.56 1.93 6.55
Noc / Est 7.56 1.93 6.55
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 17

Error Fl-bg Fl-fg Fl-all
All / All 5.92 0.97 5.40
All / Est 5.92 0.97 5.40
Noc / All 5.02 0.97 4.50
Noc / Est 5.02 0.97 4.50
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 18

Error Fl-bg Fl-fg Fl-all
All / All 15.90 99.96 55.84
All / Est 15.90 99.96 55.84
Noc / All 11.03 99.86 35.53
Noc / Est 11.03 99.86 35.53
This table as LaTeX

Input Image

Flow Result

Flow Error


Test Image 19

Error Fl-bg Fl-fg Fl-all
All / All 4.16 2.10 3.93
All / Est 4.16 2.10 3.93
Noc / All 3.41 2.10 3.22
Noc / Est 3.41 2.10 3.22
This table as LaTeX

Input Image

Flow Result

Flow Error




eXTReMe Tracker