Command Line arguments - asolis/vivaTracker GitHub Wiki

Name

vivaTracker   executes a tracking method using a video sequence and outputs the results

Synopsis

vivaTracker [-h] [-m=method] [-o outputfile] [-n] [-g=ground-truth] [-v filename | -v folder] [-p] sequence

Description

Execute the tracking method specified with the option -m over the video sequence. If no method is specified the default one will be the skcf algorithm. The sequence could be a video file, a folder containing images, a dataset sequence (e.g., vot2013/bolt or vot2013\bolt in windows) or url pointing to a video (e.g., http://urltoyourvideo.com/video.mov). If there is a file containing the ground truth under the same folder of the sequence (i.e., file with the filename "groundtruth.txt"), the vivaTracker executable will output the ground truth along the trackers output. The existing ground truth will be used to initialize the tracker for the first frame. If no ground truth is found the user will be able to select a rectangular region to track in the video with two clicks (i.e., first click defines the top left corner and second click the bottom right corner). The 'SPACE' key in your keyboard can be used to pause/play the current sequence, allowing a better selection of the target. Optional outputs can be specified to store the visual results in a video or image sequence (i.e., -v).

The following options are available:

Usage: vivaTracker [params] sequence 

	-g, --groundtruth
		specify groundtruth file
	-h, --help (value:true)
		print this message
	-m, --method (value:skcf)
		tracking method: kcf, kcf2, skcf, ncc, opentld, struck, ...
	-n, --no
		not display gui window
	-o, --output
		filename for tracking results
	-p, --pause
		start sequence paused
	-v, --video
		output video filename / folder for images output

	sequence
		url, file, folder, vot_sequence, sequence

Displaying Individual Tracking Methods Options

To display individual tracking method options you need to specify the tracking method (i.e., -m=method) and the -? option. For example, to list the details and options the skcf algorithm

$./vivaTracker -m=skcf -?

Andrés Solís Montero, Jochen Lang, Robert Laganiere (sKCF): 
Scalable Kernel Correlation Filter with Sparse Feature Integration. 2015

Usage: skcf [params] 

	-?, --usage (value:true)
		print this message
	-f, --feat (value:fhog)
		feature type: fhog, gray, rgb, hsv, hls
	-s, --scale
		turn on scale estimation
	-t, --type (value:g)
		correlation type: g(gaussian), p(polynomial), l(linear)

Using a video file (with tracking method kcf)

$./vivaTracker -m=kcf video.avi 

For a complete list of of available methods check the here.

Using a folder containing a sequence of images (with tracking method kcf2)

$./vivaTracker -m=kcf2 /path/to/img/folder/

For a complete list of of available methods check the here.

Using a sequence from a dataset

If we would like to run the opentld method with the sequence ball2 from dataset vot2015

Mac / Unix

$./vivaTracker -m=opentld vot2015/ball2

Windows

vivaTracker.exe -m=opentld vot2015\ball2

The dataset vot2015 should be already downloaded into the sequences folder. Check how to build the project for more details about the datasets.

Outputting the tracking results for a sequence

If we would like to run a tracker algorithm over a sequence (e.g., vot2015/ball2 or vot2015/ball2 in Windows) and gather the annotated tracking areas for each frame. we could run the following line

$./vivaTracker -m=skcf vot2015/ball2 -o=skcf.txt

The skcf.txt file will contain a line for each video frame in the sequence with the following data:

x1, y1, x2, y2, x3, y3, x4, y4 where (x1,y1), (x2,y2), (x3,y3) and (x4,y4) are the four corner coordinates of the tracking area in clock-wise order.

If we would like just to run the algorithm and do NOT display the output window, we could add the -n option

$./vivaTracker -m=skcf vot2015/ball2 -o=skcf.txt -n

Selecting a different ground truth file for a sequence

In this example we would like to tell the program to load a specific ground truth file. Let's assume we executed the previous example and saved the results for the vot2015/ball sequence into the file skcf.txt. We could run the following:

$./vivaTracker -m=ncc vot2015/ball2 -o=ncc.txt -g=skcf.txt 

and the algorithm will load the sequence vot2015/ball2 but using the skcf.txt as the ground truth. The output of the ncc method will be stored in ncc.txt. If we would just like to remove the ground truth annotations from the video using the -g option without any value will remove any ground truth annotations. For this option, the user will need to manually select the interest area using the mouse before start tracking.

$./vivaTracker -m=ncc vot2015/ball2 -g -p

Creating a video file with the algorithms output

$./vivaTracker -m=skcf vot2015/ball2 -v movie.avi

For a complete list of of available methods check the here.