Method

FDFlowNet: Fast Optical Flow Estimation using a Deep Lightweight Network [FDFlowNet]


Submitted on 22 Jun. 2020 14:07 by
Lingtong Kong (SJTU)

Running time:0.02 s
Environment:NVIDIA GTX 1080 Ti

Method Description:
Significant progress has been made for estimating
optical flow using deep neural networks. Advanced
deep models achieve accurate flow estimation
often with a considerable computation complexity
and time-consuming training processes. In this
work, we present a lightweight yet effective
model for real-time optical flow estimation,
termed FDFlowNet (fast deep flownet). We achieve
better or similar accuracy on the challenging
KITTI and Sintel benchmarks while being about 2
times faster than PWC-Net. This is achieved by a
carefully-designed structure and newly proposed
components. We first introduce an U-shape network
for constructing multi-scale feature which
benefits upper levels with global receptive field
compared with pyramid network. In each scale, a
partial fully connected structure with dilated
convolution is proposed for flow estimation that
obtains a good balance among speed, accuracy and
number of parameters compared with sequential
connected and dense connected structures.
Experiments demonstrate that our model achieves
state-of-the-art performance while being fast and
lightweight.
Parameters:
FDFlowNet
Latex Bibtex:
@inproceedings{kong2020fdflownet,
author={Lingtong Kong and Jie Yang},
booktitle={IEEE International Conference on
Image Processing (ICIP)},
title={FDFlowNet: Fast Optical Flow Estimation
using a Deep Lightweight Network},
year={2020},}

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 5.28 % 10.34 % 0.8 px 1.5 px
3 pixels 3.19 % 7.17 % 0.8 px 1.5 px
4 pixels 2.29 % 5.58 % 0.8 px 1.5 px
5 pixels 1.79 % 4.57 % 0.8 px 1.5 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 4.00 % 7.74 % 0.6 px 1.0 px
3 pixels 1.98 % 5.31 % 0.6 px 1.0 px
4 pixels 1.34 % 4.10 % 0.6 px 1.0 px
5 pixels 1.03 % 3.23 % 0.6 px 1.0 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 4.55 % 7.64 % 0.7 px 0.9 px
3 pixels 2.65 % 5.02 % 0.7 px 0.9 px
4 pixels 1.88 % 4.02 % 0.7 px 0.9 px
5 pixels 1.40 % 3.29 % 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 4.10 % 10.42 % 0.6 px 1.1 px
3 pixels 2.21 % 6.98 % 0.6 px 1.1 px
4 pixels 1.02 % 4.91 % 0.6 px 1.1 px
5 pixels 0.62 % 3.99 % 0.6 px 1.1 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 3.09 % 8.19 % 0.5 px 0.8 px
3 pixels 1.49 % 5.45 % 0.5 px 0.8 px
4 pixels 0.77 % 3.22 % 0.5 px 0.8 px
5 pixels 0.49 % 2.34 % 0.5 px 0.8 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 1.08 % 3.58 % 0.4 px 0.5 px
3 pixels 0.44 % 1.58 % 0.4 px 0.5 px
4 pixels 0.34 % 0.91 % 0.4 px 0.5 px
5 pixels 0.30 % 0.71 % 0.4 px 0.5 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 4.54 % 10.65 % 0.7 px 2.1 px
3 pixels 2.45 % 6.84 % 0.7 px 2.1 px
4 pixels 1.59 % 5.12 % 0.7 px 2.1 px
5 pixels 1.22 % 4.28 % 0.7 px 2.1 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 4.95 % 10.23 % 0.6 px 0.9 px
3 pixels 3.04 % 6.62 % 0.6 px 0.9 px
4 pixels 2.19 % 4.70 % 0.6 px 0.9 px
5 pixels 1.52 % 3.01 % 0.6 px 0.9 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 7.19 % 13.42 % 0.9 px 1.6 px
3 pixels 3.87 % 9.15 % 0.9 px 1.6 px
4 pixels 2.28 % 6.84 % 0.9 px 1.6 px
5 pixels 1.88 % 5.84 % 0.9 px 1.6 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 4.23 % 12.42 % 0.8 px 1.6 px
3 pixels 2.23 % 8.32 % 0.8 px 1.6 px
4 pixels 1.67 % 6.51 % 0.8 px 1.6 px
5 pixels 1.36 % 5.39 % 0.8 px 1.6 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 5.02 % 11.32 % 0.8 px 1.6 px
3 pixels 2.76 % 7.71 % 0.8 px 1.6 px
4 pixels 1.85 % 5.93 % 0.8 px 1.6 px
5 pixels 1.23 % 4.83 % 0.8 px 1.6 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 5.96 % 14.44 % 0.7 px 1.4 px
3 pixels 3.25 % 10.68 % 0.7 px 1.4 px
4 pixels 1.93 % 8.13 % 0.7 px 1.4 px
5 pixels 1.39 % 6.57 % 0.7 px 1.4 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 4.32 % 7.80 % 0.6 px 1.0 px
3 pixels 2.36 % 4.94 % 0.6 px 1.0 px
4 pixels 1.35 % 3.74 % 0.6 px 1.0 px
5 pixels 0.94 % 3.16 % 0.6 px 1.0 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 10.73 % 16.31 % 1.5 px 2.6 px
3 pixels 8.02 % 13.88 % 1.5 px 2.6 px
4 pixels 6.97 % 12.68 % 1.5 px 2.6 px
5 pixels 6.24 % 11.73 % 1.5 px 2.6 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 6.03 % 10.83 % 0.9 px 1.6 px
3 pixels 3.93 % 8.43 % 0.9 px 1.6 px
4 pixels 3.00 % 6.54 % 0.9 px 1.6 px
5 pixels 2.49 % 5.04 % 0.9 px 1.6 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 2.69 % 9.19 % 0.5 px 0.9 px
3 pixels 1.37 % 6.13 % 0.5 px 0.9 px
4 pixels 0.98 % 4.51 % 0.5 px 0.9 px
5 pixels 0.77 % 3.39 % 0.5 px 0.9 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 8.98 % 16.92 % 0.9 px 2.0 px
3 pixels 5.66 % 11.48 % 0.9 px 2.0 px
4 pixels 4.12 % 8.67 % 0.9 px 2.0 px
5 pixels 2.67 % 6.16 % 0.9 px 2.0 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 2.85 % 5.89 % 0.5 px 0.7 px
3 pixels 1.10 % 3.13 % 0.5 px 0.7 px
4 pixels 0.67 % 1.79 % 0.5 px 0.7 px
5 pixels 0.47 % 0.96 % 0.5 px 0.7 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 7.13 % 13.16 % 0.9 px 2.1 px
3 pixels 4.67 % 9.71 % 0.9 px 2.1 px
4 pixels 3.39 % 7.90 % 0.9 px 2.1 px
5 pixels 2.67 % 6.95 % 0.9 px 2.1 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 0.50 % 2.06 % 0.3 px 0.4 px
3 pixels 0.29 % 0.83 % 0.3 px 0.4 px
4 pixels 0.17 % 0.43 % 0.3 px 0.4 px
5 pixels 0.05 % 0.23 % 0.3 px 0.4 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 1.46 % 4.03 % 0.4 px 0.6 px
3 pixels 0.81 % 2.54 % 0.4 px 0.6 px
4 pixels 0.55 % 1.58 % 0.4 px 0.6 px
5 pixels 0.42 % 1.09 % 0.4 px 0.6 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot





eXTReMe Tracker