Auto Classification¶

Auto Classification description

Functions

void *rf_auto_classifier_init(const char *models_base_path, const char *model_params, rf_error_type *error)¶

Initializes classifier.

Return

classifier handle

Parameters

models_base_path: path to the models base directory or to models compound package file
model_params: settings, can be NULL to use defaults
error: output parameter to return an error details

void rf_auto_classifier_close(void *classifier_handle, rf_error_type *error)¶

Deletes classifier.

Parameters

classifier_handle: classifier handle
error: output parameter to return an error details

void rf_auto_classifier_classify_whole(void *classifier_handle, void *image_handle, const char *assigned_class_id, char **out_class_id, int max_stirng_len_out_class_id, float *out_confidence, int *top_k, rf_error_type *error)¶

Classify whole image

Parameters

classifier_handle: classifier handle
image_handle: image handle (use image API to create)
assigned_class_id: the class id that was assigned by the user (used as training data after collecting a bunch)
out_class_id: output parameter that returns a list of class ids (list should be preallocated to at least *top_k size, and each string in the list should be preallocated to at least of max_stirng_len_out_class_id size)
max_stirng_len_out_class_id: maximun preallocated length of strings in out_class_id)
out_confidence: output parameter that returns a corresponding list of confidences (should be preallocated to at least *top_k size)
top_k: input/output parameter which indicates the k top classes and confidences to return, on the output holds the value of the actual len of out_class_id and out_confidence
error: output parameter to return an error details

void rf_auto_classifier_classify(void *classifier_handle, void *image_handle, const char *assigned_class_id, int ymin, int xmin, int ymax, int xmax, char **out_class_id, int max_stirng_len_out_class_id, float *out_confidence, int *top_k, rf_error_type *error)¶

Classify an bounding box area in the image

Parameters

classifier_handle: classifier handle
image_handle: image handle (use image API to create)
assigned_class_id: the class id that was assigned by the user (used as training data after collecting a bunch)
ymin: the top y position of the bounding box where the object is located
xmin: the left x position of the bounding box where the object is located
ymax: the bottom y position of the bounding box where the object is located
xmax: the right x position of the bounding box where the object is located
out_class_id: output parameter that returns a list of class ids (list should be preallocated to at least *top_k size, and each string in the list should be preallocated to at least of max_stirng_len_out_class_id size)
max_stirng_len_out_class_id: maximun preallocated length of strings in out_class_id)
out_confidence: output parameter that returns a corresponding list of confidences (should be preallocated to at least *top_k size)
top_k: input/output parameter which indicates the k top classes and confidences to return, on the output holds the value of the actual len of out_class_id and out_confidence
error: output parameter to return an error details