Detection of Moving Objects with Non-stationary Cameras in 5.8ms: Bringing Motion Detection to Your Mobile Device
This Git repository is an implementation of the paper "Detection of Moving Objects with Non-stationary Cameras in 5.8ms: Bringing Motion Detection to Your Mobile Device," Yi et al, CVPRW 2013. These codes should reproduce the results presented in the paper, with a bit of tuning on the parameters. The results may differ a bit, as the variance update equation was modified from the one used to produce results of the paper. However, they should not differ significantly.
-
The code that we ditributed earlier through e-mail had an issue that it only gave good results with MS compiler. There was a bug that abs function was used instead of fabs (the floating point version)
-
This repository is not finalized. The current version cannot save results as video, and is also using a very old open cv style. We intend to fix these. Also, there are some redundant relics from old codes.
-
Again, this repository is build from a very old backup I had. You can go ahead an try, as the detection results won't change, but bare in mind that there might be compiler related issues.
-
This repository requires OpenCV 2.4.X
-
Python version provided by @alehdaghi. Thank you!
Simply use CMake and target the output directory as "build" in the same level as "src". In command line this would be (from the project root folder)
project_root >> mkdir build
project_root >> cd build
project_root/build >> cmake ..
project_root/build >> make
Once it is built, you can try running
project_root/build >> ./fastMCD ../data/woman.mp4 0
What you mostly need are only two files:
src/params.hpp
src/prob_model.hpp
Usage is pretty straightforward. Simply init, motion compensate, and update.
Copyright (c) 2016 Kwang Moo Yi.
All rights reserved.
This software is strictly for non-commercial use only. For commercial use, please contact me at kwang.m.yi_at_gmail.com. Also, when used for academic purposes, please cite the paper "Detection of Moving Objects with Non-stationary Cameras in 5.8ms: Bringing Motion Detection to Your Mobile Device," Yi et al, CVPRW 2013 Redistribution and use for non-commercial purposes in source and binary forms are permitted provided that the above copyright notice and this paragraph are duplicated in all such forms and that any documentation, advertising materials, and other materials related to such distribution and use acknowledge that the software was developed by the Perception and Intelligence Lab, Seoul National University. The name of the Perception and Intelligence Lab and Seoul National University may not be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY WARRANTIES. USE AT YOUR OWN RISK!
The test video is the woman dataset from the FragTrack Website. If you use it, please cite, Amit Adam, Ehud Rivlin, Ilan Shimshoni: "Robust Fragments-based Tracking using the Integral Histogram." Proc. CVPR 2006, pp. 798-805