dat's Introduction

DAT

This is a C++ reimplementation of algorithm presented in "In Defense of Color-based Model-free Tracking." . Since the source code contains no multiscale module, we just retain distractor-aware object module which is rather fast. For more info, visit the autor's webpage!.

Usage

Simple test can be compiled like:

cmake
make
.\DAT

For test the vot2015 challenge, please download the dataset at first and then unzip to ${DAT}.

cmake
make
.\DAT vot2015

Speed

Test in Windows with Intel i7-4790

Test in ubuntu 16.04 with Intel i7-6700

Reference

@INPROCEEDINGS{possegger15a,
  author = {Horst Possegger and Thomas Mauthner and Horst Bischof},
  title = {In Defense of Color-based Model-free Tracking},
  booktitle = {Proc. IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  year = {2015}
}

dat's People

Contributors

Stargazers

Watchers

dat's Issues

Explanation needed

	prob_map = cv::Mat(frame.size(), CV_32FC1);
	cv::Mat frame_bin;
	cv::LUT(frame, bin_mapping, frame_bin);

	float *p_prob_map = prob_map.ptr<float>(0);
	cv::MatIterator_<cv::Vec3b> it, end;
	for (it = frame_bin.begin<cv::Vec3b>(), end = frame_bin.end<cv::Vec3b>(); it != end; ++it)
	{
		*p_prob_map++ = prob_lut.at<float>((*it)[0], (*it)[1], (*it)[2]);
	}

Hi
Thanks for providing such a beautifully written code. I would be very thankful if you could explain what the above lines of code are doing.
Thanks 👍

Recommend Projects