CVBpy 15.0
SearchPredictor Class Reference

Predictor that may be used for searching objects. More...

Inherits PredictorBase.

Public Member Functions

Tuple[cvb.polimago.SearchResult, int] grid_search (self, cvb.Image img, cvb.Rect aoi, float grid_step, float threshold, float locality)
 Perform a grid search. More...
 
Tuple[bool, cvb.polimago.SearchResult, int, List[cvb.polimago.SearchResult]] inspect (self, cvb.Image img, cvb.polimago.SearchResult res, int search_depth)
 Carries out the operation that grid_search executes for a grid point, starting at the perspective and position defined by the initial value of the parameter SearchResult. More...
 
cvb.Image search_result_to_image (self, cvb.Image source_image, cvb.polimago.SearchResult res)
 Create a visual representation of a search result. More...
 
- Public Member Functions inherited from PredictorBase
bool is_compatible (self, cvb.Image img, cvb.Point2D pos)
 Verify the compatibility of a CVB image with this classifier. More...
 
- Public Member Functions inherited from PolimagoFactoryCreatedObject
None save (self, str file_name)
 Save this object into a file. More...
 

Properties

 extraction_radius = property
 float: The radius for extracting positive search instances. More...
 
 feature_resolution_rest = property
 int: The feature resolution to be used for the third and later classification steps.
 
 feature_resolution_step_1_and_2 = property
 int: The feature resolution to be used for the first two classification steps.
 
 invariances = property
 int: The invariances that have been trained on this classifier (see cvb.polimago.InvarianceType).
 
 max_num_results = property
 int: The maximum number of results that can be extracted in a GridSearch operation. More...
 
 num_classification_steps = property
 int: The number of classification steps defined during training.
 
 rotation_range = property
 cvb.AngleRange: The range of angles that was covered during classifier training.
 
 sample_size = property
 int: The sample size that has been used in each training set.
 
 scale_range = property
 cvb.NumberRange: The range of scale factors that was covered during classifier training.
 
- Properties inherited from PolimagoFactoryCreatedObject
 correction = property
 cvb.Point2D: Correction factors in X and Y direction required to rescale the input images for projection onto the retina.
 
 feature_window_extent = property
 cvb.Rect: The feature window extent that has been used during classifier training.
 
 file_name = property
 str: Name of the file the object has been loaded from (or empty string if the object was not loaded).
 
 image_planes = property
 int: Intercept weight that has been used for generating this object.
 
 interpolation = property
 int: Interpolation setting used for generating this object (see cvb.polimago.InterpolationType).
 
 lambda_ = property
 float: Regularization value that has been used for generating this object.
 
 offset = property
 float: Intercept weight that has been used for generating this object.
 
 preprocessing = property
 str: Preprocessing code with which this object was generated.
 
 retina_size = property
 cvb.Size2D: Size of the 'Retina' in pixels. More...
 

Detailed Description

Predictor that may be used for searching objects.

Load a saved Polimago search predictor from a file.

Parameters

file_name : str Name of the file to be loaded.

Member Function Documentation

◆ grid_search()

Tuple[cvb.polimago.SearchResult, int] grid_search (   self,
cvb.Image  img,
cvb.Rect  aoi,
float  grid_step,
float  threshold,
float  locality 
)

Perform a grid search.

If two results that exceed the quality requirement defined by the threshold parameter are spaced less than 'locality' pixels apart (in either x or y direction) then the result with the lower quality will be eliminated from the result list.

Parameters

img : cvb.Image Image to search in.

aoi : cvb.Rect AOI inside the image to search in.

grid_step : float Spacing of the grid points in units of the feature window width/height.

threshold : float Minimum quality a result must have for being reported.

locality : float Minimum distance between the results that this SearchPredictor found.

Returns

Tuple[cvb.polimago.SearchResult, int] Tuple containing list of results that this SearchPredictor found in the area of interest and number of calls to the predictor's classification routine that have been carried out in the GridSearch call.

◆ inspect()

Tuple[bool, cvb.polimago.SearchResult, int, List[cvb.polimago.SearchResult]] inspect (   self,
cvb.Image  img,
cvb.polimago.SearchResult  res,
int  search_depth 
)

Carries out the operation that grid_search executes for a grid point, starting at the perspective and position defined by the initial value of the parameter SearchResult.

When the function returns successfully, it will return SearchResult containing the final perspective and result quality. For additional information please see the tutorial on search functions of Polimago.

Parameters

img : cvb.Image Image to work on.

res : cvb.polimago.SearchResult Search Result to start with.

search_depth : int Maximum number of steps to calculate (upon exit: number of steps actually calculated).

Returns

Tuple[bool, cvb.polimago.SearchResult, int, List[cvb.polimago.SearchResult]] Tuple containing success (if predictor found a result starting at the position defined through res parameter), final SearchResult, number of steps, and list of SearchResults leading from initial to final SearchResult value.

◆ search_result_to_image()

cvb.Image search_result_to_image (   self,
cvb.Image  source_image,
cvb.polimago.SearchResult  res 
)

Create a visual representation of a search result.

Parameters

source_image : cvb.Image Image to create the representation from. This should be the image on which the result was found with this classifier.

res : cvb.polimago.SearchResult Search result (should match the image and this classifier).

Returns

cvb.Image Visual representation of the search result.

Property Documentation

◆ extraction_radius

extraction_radius = property
static

float: The radius for extracting positive search instances.

The unit size is the size of the feature window, i.e. a value of 0.5 means that positive samples can be extracted from withing a range of +/- Feature Window Width and +/- Feature Window Height around each trained instance. Smaller values will necessitate a finer search grid but may yield better results on difficult search tasks.

◆ max_num_results

max_num_results = property
static

int: The maximum number of results that can be extracted in a GridSearch operation.

Increasing this value will increase the amount of memory allocated by this SearchPredictor object.