qingchenbenpao/epipolarscore

[WACV-2020] Exploiting Geometric Constraints on Dense Trajectories for Motion Saliency

EpipolarScore

This repository is part of public implementation of our "Exploiting Geometric Constraints on Dense Trajectories for Motion Saliency" paper. This repository contains Optical Flow & Epipolar Score Computation code. You can check the main repository here.

Installations:

This source code is based on MATLAB framework and tested on Ubuntu 16.04 with MATLAB 2016b.

Instructions:

1) Clone the repository

git clone https://github.com/mfaisal59/EpipolarScore.git

2) Download Dataset

Download and unpack the DAVIS 2016 dataset and as well as the evaluatio code from https://davischallenge.org/davis2016/code.html

3) Compute Optical Flow

The optical flow is based on Full Flow Method (https://cqf.io/fullflow/). To compute the optical flow for DAVIS Dataset, run the following script:

cd ./Full_Flow_Source_Code/
run davisBatch.m file
#modify the path to DAVIS dataset directory

4) Compute Epipolar Score

To compute the Epipolar Score, modify the paths in 'testDAVIS.m' file and run:

cd ./EpipolarScoreMain/
run testDAVIS.m script
#modify the path to DAVIS dataset, forward and backward optical flow directory

5) Convert Flow to X-Y Displacement Images

cd ./EpipolarScoreMain/
run flow2Displacement.m script
#modify the paths to DAVIS dataset, forward and backward optical flow directory

6) Generate Motion Images

cd ./EpipolarScoreMain/
run generateMotionImages.m script
#modify the paths to DAVIS dataset and Optical Flow directory

BIBTEX:

@article{DBLP:journals/corr/abs-1909-13258,
  author    = {Muhammad Faisal and
               Ijaz Akhter and
               Mohsen Ali and
               Richard I. Hartley},
  title     = {Exploiting Geometric Constraints on Dense Trajectories for Motion
               Saliency},
  journal   = {CoRR},
  volume    = {abs/1909.13258},
  year      = {2019},
  url       = {http://arxiv.org/abs/1909.13258}
}