Command Line Options

The annotate tool

Name:

   annotate  -- create ground-truth files with polygon annotations

Synopsis:

   annotate [-h|help] [-m={p,r,a}] [-W] [-H] [-r] [-t={e,d,n}] [-o] sequence

Description:

   Annotate is a tool to create ground-truth file annotations for a sequence. A sequence could be a video file or a set of images. The Annotate tool provides next frame prediction of the current selected target using our [sKCF]() tracking solution. 

The following options are available:

      sequence            :  it could be a video file, a folder containing images... (see vivalib library for more options)

      -h|help             :  displays the command line help

      -m|mode={p,r,a}     :  selects the annotation mode, the supported choices are p (polygon), r(rotated rectangle), a(axis align rectangle) 

      -W|width=value      :  rescales the input to have _value_ sized width while keeping the aspect ratio. The default value is -1 (i.e., uses the original input's width).

      -H|height=value     :  rescales the input to have _value_ sized height while keeping the aspect ratio. The default value is -1 (i.e., uses the original input's height).

      -r|ratio=value      : forces the ratio height/width of the annotations while using method p and r.

      -t|track={e,d,n}    : enables/disable tracking mode. The 'e'  enable option will propose a new position for each annotation in the next frame using our sKCF algorithm. The 'd' disable option will remove the tracking capabilities and will keep the annotations position from the previous frame. The 'n' new option, will not copy the annotations from previous frame to new frames. This mode is useful when your are trying to annotated a sequence that is not continuous (i.e., a set of images not part of a video).

       -o|output          : filename for  the annotation results

The extract tool

Name:

   extract  -- extract polygon annotations from a sequence by creating independent files.

Synopsis:

   extract [-h|help] [-g] [-o] [-p] [-m] [-s] sequence

Description:

   Extract is a tool to extract ground-truth annotations for a sequence. The annotations follow the format created by the **annotate** tool. A sequence could be a video file or a set of images.  

The following options are available:

      sequence            :  it could be a video file, a folder containing images... (see vivalib library for more options)

      -h|help             :  displays the command line help

      -g                  :  specifies the ground truth file with the annotations 

      -o                  :  folder name to store the output annotations (default value: "output")

      -p                  :  pattern to specify the filename to create (i.e., %06d%03d.png, will generate filenames like this 000001001.png, 000002001.png). The first '%d' representes the frame number and the second '%d' the identifier of the annotation inside that frame. They are both 0-index. 

      -m|masked           : enable this options when you would like to remove the background out of the polygonal annotation. 

       -s                 : start the video sequence in pause mode. Hit 'n' key to go the next frame or 'SPACE' to play/pause the video sequence.