Method

FastFlowNet: A Lightweight Network for Fast Optical Flow Estimation [FastFlowNet]
https://github.com/ltkong218/FastFlowNet

Submitted on 14 Sep. 2020 05:42 by
Lingtong Kong (SJTU)

Running time:0.01 s
Environment:NVIDIA GTX 1080 Ti

Method Description:
Dense optical flow estimation plays a key role in
many robotic vision tasks. It has been predicted
with satisfying accuracy than traditional methods
with advent of deep learning. However, current
networks often occupy large number of parameters
and require heavy computation costs. These
drawbacks have hindered applications on power- or
memory-constrained mobile devices. To deal with
these challenges, in this paper, we dive into
designing efficient structure for fast and
accurate optical flow prediction. Our proposed
FastFlowNet works in the well-known coarse-to-fine
manner with innovations of head enhanced pooling
pyramid (HEPP) feature extractor, center dense
dilated correlation (CDDC) layer and efficient
shuffle block decoder (SBD). Experiments on both
synthetic Sintel and real-world KITTI datasets
demonstrate the effectiveness of proposed
approaches, which consumes only 1/10 computation
of comparable networks to get 90% of their
performance.
Parameters:
FastFlowNet
Latex Bibtex:
@InProceedings{Kong_2021_ICRA,
author={Kong, Lingtong and Shen, Chunhua and Yang,
Jie},
title={FastFlowNet: A Lightweight Network for Fast
Optical Flow Estimation},
booktitle={2021 IEEE International Conference on
Robotics and Automation (ICRA)},
year={2021}
}

Detailed Results

This page provides detailed results for the method(s) selected. For each of the first 20 test images, the number of erroneous pixels at all thresholds is depicted in the table. Underneath, the left input image, the disparity / end-point error map and the estimated (and interpolated) disparity / optical flow map are shown. The error map scales linearly between 0 (black) and >=5 (white) pixels error. Red denotes all occluded pixels, falling outside the image boundaries. The false color map is scaled to the largest ground truth disparity / flow value.

Test Set Average

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 6.06 % 11.71 % 0.9 px 1.8 px
3 pixels 3.78 % 8.34 % 0.9 px 1.8 px
4 pixels 2.79 % 6.58 % 0.9 px 1.8 px
5 pixels 2.23 % 5.47 % 0.9 px 1.8 px
This table as LaTeX

Reflective Regions

This table as LaTeX

Test Image 0

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 7.59 % 12.04 % 0.9 px 1.5 px
3 pixels 5.01 % 8.28 % 0.9 px 1.5 px
4 pixels 4.07 % 6.65 % 0.9 px 1.5 px
5 pixels 3.60 % 5.77 % 0.9 px 1.5 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 1

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 5.02 % 8.71 % 0.7 px 0.9 px
3 pixels 2.94 % 6.17 % 0.7 px 0.9 px
4 pixels 2.12 % 4.78 % 0.7 px 0.9 px
5 pixels 1.55 % 3.66 % 0.7 px 0.9 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 2

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 6.70 % 13.19 % 0.7 px 1.3 px
3 pixels 3.89 % 9.35 % 0.7 px 1.3 px
4 pixels 2.53 % 7.05 % 0.7 px 1.3 px
5 pixels 1.49 % 5.35 % 0.7 px 1.3 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 3

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 4.65 % 11.08 % 0.6 px 0.9 px
3 pixels 2.73 % 7.49 % 0.6 px 0.9 px
4 pixels 1.49 % 4.70 % 0.6 px 0.9 px
5 pixels 0.90 % 3.07 % 0.6 px 0.9 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 4

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 1.41 % 3.57 % 0.5 px 0.6 px
3 pixels 0.55 % 1.23 % 0.5 px 0.6 px
4 pixels 0.43 % 0.66 % 0.5 px 0.6 px
5 pixels 0.40 % 0.51 % 0.5 px 0.6 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 5

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 7.02 % 12.53 % 1.0 px 1.9 px
3 pixels 4.09 % 8.22 % 1.0 px 1.9 px
4 pixels 3.21 % 6.77 % 1.0 px 1.9 px
5 pixels 2.74 % 5.88 % 1.0 px 1.9 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 6

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 6.20 % 11.47 % 0.7 px 1.0 px
3 pixels 3.51 % 7.55 % 0.7 px 1.0 px
4 pixels 2.23 % 5.23 % 0.7 px 1.0 px
5 pixels 1.66 % 3.64 % 0.7 px 1.0 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 7

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 7.43 % 14.62 % 0.9 px 2.4 px
3 pixels 5.16 % 10.78 % 0.9 px 2.4 px
4 pixels 3.89 % 8.60 % 0.9 px 2.4 px
5 pixels 3.08 % 7.23 % 0.9 px 2.4 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 8

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 5.38 % 14.19 % 0.9 px 1.7 px
3 pixels 2.79 % 9.74 % 0.9 px 1.7 px
4 pixels 2.02 % 7.54 % 0.9 px 1.7 px
5 pixels 1.64 % 6.09 % 0.9 px 1.7 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 9

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 6.84 % 13.98 % 0.9 px 1.8 px
3 pixels 4.23 % 10.12 % 0.9 px 1.8 px
4 pixels 2.97 % 7.93 % 0.9 px 1.8 px
5 pixels 2.07 % 6.19 % 0.9 px 1.8 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 10

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 4.87 % 12.30 % 0.7 px 1.1 px
3 pixels 2.48 % 8.49 % 0.7 px 1.1 px
4 pixels 1.52 % 6.40 % 0.7 px 1.1 px
5 pixels 1.08 % 4.93 % 0.7 px 1.1 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 11

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 6.64 % 12.36 % 0.7 px 1.5 px
3 pixels 3.54 % 8.08 % 0.7 px 1.5 px
4 pixels 2.07 % 5.66 % 0.7 px 1.5 px
5 pixels 1.31 % 4.14 % 0.7 px 1.5 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 12

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 10.37 % 16.60 % 1.6 px 3.1 px
3 pixels 7.73 % 13.44 % 1.6 px 3.1 px
4 pixels 6.77 % 12.28 % 1.6 px 3.1 px
5 pixels 6.33 % 11.66 % 1.6 px 3.1 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 13

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 5.78 % 10.64 % 1.0 px 1.7 px
3 pixels 4.30 % 8.19 % 1.0 px 1.7 px
4 pixels 3.61 % 6.55 % 1.0 px 1.7 px
5 pixels 3.19 % 5.33 % 1.0 px 1.7 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 14

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 4.12 % 12.56 % 0.6 px 1.3 px
3 pixels 2.01 % 8.87 % 0.6 px 1.3 px
4 pixels 1.42 % 7.16 % 0.6 px 1.3 px
5 pixels 1.19 % 5.94 % 0.6 px 1.3 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 15

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 11.12 % 19.02 % 1.1 px 2.5 px
3 pixels 7.41 % 13.72 % 1.1 px 2.5 px
4 pixels 5.43 % 10.31 % 1.1 px 2.5 px
5 pixels 4.24 % 8.01 % 1.1 px 2.5 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 16

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 3.72 % 8.28 % 0.6 px 0.8 px
3 pixels 1.80 % 5.41 % 0.6 px 0.8 px
4 pixels 1.12 % 3.86 % 0.6 px 0.8 px
5 pixels 0.66 % 2.73 % 0.6 px 0.8 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 17

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 7.94 % 14.63 % 0.9 px 2.9 px
3 pixels 5.14 % 10.60 % 0.9 px 2.9 px
4 pixels 3.76 % 8.62 % 0.9 px 2.9 px
5 pixels 2.91 % 7.36 % 0.9 px 2.9 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 18

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 0.71 % 4.27 % 0.4 px 0.5 px
3 pixels 0.38 % 2.91 % 0.4 px 0.5 px
4 pixels 0.25 % 1.91 % 0.4 px 0.5 px
5 pixels 0.13 % 1.28 % 0.4 px 0.5 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 19

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 1.52 % 4.43 % 0.4 px 0.6 px
3 pixels 0.94 % 2.87 % 0.4 px 0.6 px
4 pixels 0.73 % 1.80 % 0.4 px 0.6 px
5 pixels 0.65 % 1.30 % 0.4 px 0.6 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot





eXTReMe Tracker