VPLAN.FINDER (cam, mode, dmode, arg1, arg2, ambiguity) $fmods[], $bmods[]

VPLAN.FINDER-I

Function

Set up the type of "planning" used by the ObjectFinder when locating models.

Usage Considerations

Adept recommends that you allow disambiguation between models by leaving the ambiguity parameter at its default value (0).

NOTE: The decision to perform intramodel symmetry disambiguation is always determined by the value set for the VTRAIN.FINDER program instruction's symmetry override parameter (mode 1, arg3). See the description of VTRAIN.FINDER (cam, mode, dmode, arg, arg2, arg3) $model_name, ibr = value, value2, value3 for details.

Parameter s

cam Real-valued expression indicating the virtual camera number to use for associating "planning" with and to read parameters from.

mode Real-valued expression indicating the planning mode:
0 = Basic initial planning
1 = Quick initial planning (for multi-instance training)

dmode Not used.

arg1 Not used.

arg2 Not used.

ambiguity Real-valued expression indicating the type of disambiguation:
0 = Perform intermodel disambiguation (default)
1 = Disable intermodel disambiguation

$fmods[] An array of foreground models. Each list must contain names of valid, trained models, starting with index [0] and terminating with a null string as the last name.

$bmods[] An array of background models. Each list must contain names of valid, trained models, starting with index [0] and terminating with a null string as the last name.

Details

Foreground models are those for which you would like the ObjectFinder to return locations. Background models are those that may be present, but which you are not interested in locating. Specifying these, when applicable, will help planning tell which features of the foreground models are best for finding proposals that are not ambiguous with respect to other similar-looking objects in the scene.

Description of planning modes

Mode 0 planning takes the lists of models and does all the preparation work needed for recognition ahead of time, to allow for the fastest possible recognition times. Planning is usually fairly fast but can take many seconds to compute and many Kb to store (from V+ program memory) if the total number of model pairs is large.

Mode 1 planning is faster than Mode 0 planning but causes recognition to be slower (however, you may achieve more reliable results using Mode 0).

NOTE: Mode 1 is primarily intended for implementing learning, not for normal recognition.

The differences between Mode 1 and Mode 0 are caused by several factors, most significant of which is that confirmation is not planned for in Mode 1. Therefore, during recognition, all proposals go on to verification. This can make Mode 1 planning costly since verification is a time-consuming operation.

Input parameters using virtual cameras

The following settings are required:

V.MIN.LEN[vc] = 20 Percentage of features (by length) to ignore.

For Mode 0 (normal planning), V.MIN.LEN is ignored.

For Mode 1 (quick planning), V.MIN.LEN[vc] is the number of features to use for planning. For example, a value of 20 means to plan with the 20 longest features.

The following settings are suggested:

V.MAX.VER.DIST[vc] = 5.0

Related Keyword

VPLAN.FINDER (monitor command)

Syntax

VPUTCAL-I