Mbarivision Options

 The table below lists all the mbarivision options arranged in 3 categories: Input Reading/Formatting Options, Output Writing/Formatting Options and Special Features Options. There are other options, outside this table that pertain to the saliency toolkit. The entire list of options can be found using the mbarivision --help switch.

Option	Default	Description
Input Reading/Formatting
--input-frames=[first-last]@[delay_or_rate]	required	Input frame range and inter-frame delay or rate. The frame range can include optional specifications for the first frame (default value=0), last  frame (default=max), and/or frame step (default=1). A fixed framerate can be specified either as an inter-frame interval, with a suffix one of (s,ms,us,ns), or as a frame rate, with a suffix of Hz.
--crop-input=x1,y1,x2,y2	[0,0,0,0]	Crop input frames, or 0,0,0,0 for no cropping.
--rescale-input=<width>x<height>	[0x0]	Rescale input frames to fixed dims, or 0x0 for no rescaling
--[no]preserve-input-aspect	[no]	Preserve input frame aspect ratio if rescaling to fixed dims
--zero-number-frames=<true|false>	[false]	Force all input and output frames to have the number 000000
--in=<[type]:[spec]>	required	Reads input from one or more raster files. The leading 'raster:' may be omitted if the file has one of the known raster extensions: .pnm, .pgm, .ppm, .pbm, .pfm, .png, .jpeg, .jpg, .rgb555, .rgb565, .rgb24, .rgb32, .yuyv, .uyvy, .yuv444, .yuv422, .yuv411, .yuv420, .yuv410, .yuv444p, .yuv422p, .yuv411p, .yuv420p, .yuv410p; or, any of the above with an additional .gz or .bz2 extension, in which case the image file will be transparently decompressed. If the given filename includes a pair of hashes ('#'), then characters in between the hashes will be interpreted as a minimum numeric field width, and the frame number will be formatted as a zero-padded string to that minimum width, and will be substituted for the entire #nnn# sequence. As special cases, a single '#' is equivalent to '#6#', and '##' is equivalent to '#0#'. Thus a filename of foo#.png or foo#6#.png will cause the stream to read foo000000.png, foo000001.png, etc.; foo##.png or foo#0#.png will read foo0.png, foo1.png, etc.; and foo#3#.png will generate foo000.png, foo001.png, etc. It is important to format the filenames with a single dot '.' e.g. a filename formatted like T.f00000.ppm isn't supported, but Tf00000.ppm is okay, in this case you would pass the in argument with Tf#.ppm.
--mbari-load-events=fileName	[]	Load the event structure from a text file instead of computing it from the frames
--mbari-load-properties=fileName	[]	Load the event property vector from a text file
Output Writng/Formatting
--out=<raster|display|mpeg|none>	required	Value recomended is none.
--[no]mbari-save-results	[no]	Save intermediate results in MBARI programs to disc
--[no]mbari-display-results	[no]	Display intermediate results in MBARI programs
--mbari-mark-interesting=<None|Shape|Outline|BoundingBox>	[BoundingBox]	Way to mark interesting events in output frames of MBARI programs
--mbari-opacity=<0.0 ... 1.0>	[1.0]	Opacity of shape or outline markings of events
--[no]mbari-mark-candidate	[yes]	Mark candidates for interesting events in output frames of MBARI programs
--[no]mbari-mark-prediction	[no]	Mark the Kalman Filter's prediction for the location of an object in output frames of MBARI programs
--[no]mbari-mark-foe	[no]	Mark the focus of expansion in the output frames of MBARI programs
--[no]mbari-save-output	[no]	Save output frames in MBARI programs
--[no]mbari-display-output	[no]	Display output frames in MBARI programs
--[no]mbari-label-events	[yes]	Write event labels into the output frames
--mbari-rescale-display=<width>x<height>	[0x0]	Rescale displays to <width>x<height>, or 0x0 for no rescaling
--mbari-save-events=fileName	[]	Save the event structure to a text file
--mbari-save-properties=fileName	[]	Save the event property vector to a text file
--mbari-save-positions=fileName	[]	Save the positions of events to a text file
--mbari-save-event-num=ev1,ev1,...,evN; or: all	[]	Save video clips showing specific events
--mbari-save-event-summary=fileName	[]	Save a human readable summary of all the events to a text file
--[no]mbari-save-non-interesting-events	[yes]	If saving events, save non interesting events in addition to interesting event. Default is to not save non interesting events.
--mbari-save-events-xml=fileName	[]	Save a XML output per all events
--mbari-source-metadata=fileName	[]	Add video input source information to XML output
Special Features
-mbari-mosaic-benthic-stills	[no]	Implements good choice of options to experiment with processing still images from a still or moving camera traversing the sea bottom. EQUIVALENT TO: --mbari-saliency-dist=1 --mbari-tracking-mode=None --mbari-keep-boring-WTA-points=yes --mbari-save-non-interesting-events=yes --mbari-segment-algorithm-input-image=MaxRGB --mbari-saliency-input-image=Raw --mbari-cache-size=2 --mbari-max-WTA-points=15 --mbari-max-evolve-msec=15000 --vc-type=OIC --use-random=false --test-mode=true --oricomp-type=Steerable --mbari-cache-size=2 --use-random=false --use-older-version=false --shape-estim-mode=ConspicuityMap --ior-type=ShapeEst
-mbari-benthic-video	[no]	Implements good choice of options to experiment with processing video from a moving camera traversing the sea bottom. EQUIVALENT TO: --mbari-saliency-dist=5 --mbari-tracking-mode=NearestNeighbor --mbari-saliency-input-image=DiffMean --mbari-segment-algorithm-input-image=MaxRGB --vc-type=O:5IC --use-random=false --test-mode=true --ori-interaction=None --oricomp-type=Steerable --mbari-cache-size=15 --use-random=false --use-older-version=false --shape-estim-mode=ConspicuityMap --ior-type=ShapeEst
-mbari-midwater-video	[no]	Implements good choice of options to experiment with processing video from a moving camera traversing the midwater sea column. EQUIVALENT TO: --mbari-saliency-dist=5 --mbari-tracking-mode=KalmanFilter --mbari-saliency-input-image=DiffMean --mbari-segment-algorithm-input-image=MaxRGB --vc-type=O:5IC --use-random=false --test-mode=true --mbari-cache-size=10 --use-random=false --use-older-version=false --shape-estim-mode=ConspicuityMap --ior-type=ShapeEst
-mbari-mosaic-stills this option is not supported in pmbarivision	[no]	Implements good choice of options to experiment with still frames collected from a moving camera in mosaic form. EQUIVALENT TO: --mbari-saliency-dist=1 --mbari-tracking-mode=None --mbari-keep-boring-WTA-points=yes --boring-sm-mv=0.25 --mbari-save-non-interesting-events=yes --mbari-segment-algorithm-input-image=MaxRGB --vc-type=Variance --use-random=false --mbari-saliency-input-image=Raw --mbari-cache-size=2 --mbari-max-WTA-points=25 --mbari-max-evolve-msec=15000
-mbari-timelapse-stills this option is not supported in pmbarivision	[no]	Implements good choice of options to experiment with still frames collected from a stationary time-lapse camera. EQUIVALENT TO: --mbari-saliency-dist=1 --mbari-tracking-mode=NearestNeighbor --mbari-keep-boring-WTA-points=yes --mbari-save-non-interesting-events=yes --mbari-segment-algorithm-input-image=MaxRGB --mbari-saliency-input-image=Raw --mbari-cache-size=10 --qtime-decay=1.0 --vc-type=Variance --use-random=false --mbari-max-WTA-points=30 --mbari-max-evolve-msec=15000 --use-random=false --use-older-version=false
-mbari-timelapse-rover-stills	[no]	Implements good choice of options to experiment with time-lapse still frames collected from a benthic moving camera . EQUIVALENT TO: --mbari-saliency-dist=1 --mbari-tracking-mode=None --mbari-keep-boring-WTA-points=yes --qtime-decay=1.0 --mbari-save-non-interesting-events=yes --mbari-segment-algorithm-input-image=MaxRGB --mbari-saliency-input-image=Raw --vc-type=O:5IC --use-random=false --mbari-max-WTA-points=15 --mbari-max-evolve-msec=15000
--mbari-max-WTA-points=<int>	[20]	Maximum number of winner-take-all points to find in each frame
--mbari-max-evolve-msec=<int>	[500]	Maximum amount of time in milliseconds to evolve the brain until stopping
--mbari-cache-size=<int>	[30]	The number of frames used to compute the running average.
--mbari-tracking-mode=<KalmanFilter|BoundingBox>	[KalmanFilter]	Tracking mode used to track events between frames
--mbari-segment-algorithm=<BinaryAdaptive|AdaptiveThreshold|BackgroundCanny|HomomorphicCanny|GraphCut>	[BinaryAdaptive]	Segmentation algorithm
--mbari-segment-algorithm-input-image=<MaxRGB|Luminance>	[MaxRGB]	Segment algorithm input images type
--mbari-segment-algorithm-se-type=<benthic|midwater>	[benthic]	Jerome Mariette's segmentation algorithm structure element test
--mbari-segment-input-image=<Raw|DiffMean>	[DiffMean]	Saliency input image type
--mbari-mask-path=<file>	[]	MaskPath: path to the mask image. Important this only masks the area to look for detections - it does not mask the area from the saliency computation.
--mbari-mask-xposition=<int>	[1]	MaskXPosition: x position of the mask point of reference
--mbari-mask-yposition=<int>	[1]	MaskYPosition: y position of the mask point of reference
--mbari-mask-width=<int>	[1]	MaskWidth: mask width
--mbari-mask-height=<int>	[1]	MaskHeight: mask height
--mbari-min-event-area=<int>	[34]	The minimum area an event must be to be candidate
--mbari-max-event-area=<int>	[1000]	The maximum area an event can be, to be candidate
--mbari-min-event-frames=<int>	[1]	The minimum number of frames an event must be to be candidate
--mbari-max-event-frames=<int>	[-1]	The maximum number of frames an event can be, to be candidate. Defaults to infinite
--mbari-saliency-dist=<int>	[5]	The number of frames to delay between saliency map computations Reduce this to improve your detection rate. Warning reducing this will increase your computation time.