PROJECT: Apreture robust optical flow for event driven sensors
This code performs aperture corrected flow computation over events obtained from DVS or ATIS sensors.
The events must be written in a txt file as: N x 4 array where N = number of events
Each row must be in form:
x y timestamp polarity
** INSTALLATION **
Dependencies: Eigen3, gcc, g++
Install dependencies:
sudo apt-get install libeigen3-dev gcc g++ cmake
To install project (recommended):
mkdir _build
cd _build
cmake ..
make
To run the project:
cd _build
Type ./FARMS_Flow --help for parameters list ..
e.g: ./FARMS_Flow --width 320 --height 320 --filename ~/path/to/file/filename *note: filename is without extension
Parameter list:
--help Displays this message
--filename arg add events file name without extension (.txt)
--height arg set sensor height
--width arg set sensor width
--filtersize arg set size of neighbor for plane fitting
--inlierCheck arg set minimum number of inliers to validate plane
--numEvents arg set max number of events to process
Input event files: txt files with N rows for each event storing events as: x y t p.
Output files: with suffix FARMSOut with N rows for each event storing flow events as: x y t p globalR globalTheta localR localTheta
Please cite as: H. Akolkar, S. H. Ieng and R. Benosman, "Real-time high speed motion prediction using fast aperture-robust event-driven visual flow," in IEEE Transactions on Pattern Analysis and Machine Intelligence, doi: 10.1109/TPAMI.2020.3010468.
Send Queries to : himanshu.akolkar@gmail.com (or use the github ticketing)