 |
SqueezeBrains SDK 1.18
|
|
SqueezeBrains SDK 1.18
|
SVL parameters. More...
#include <sb.h>
Data Fields | |
| sb_fp_svl_pre_elaboration | fp_pre_elaboration |
| Callback to execute an image pre elaboration. More... | |
| sb_fp_svl_progress | fp_progress |
| The SVL calls this callback to notify the user the results of SVL. More... | |
| sb_fp_svl_command | fp_command |
| Callback called by SVL to allow the user to decide how to continue when particular situations happen, see sb_t_svl_stop_reason . More... | |
| void * | user_data |
| Pointer to data which is passed to the callbacks. More... | |
| char | project_path [512] |
| Path of the project, where the SVL will find the images. | |
| char | image_ext [64] |
| Extensions of the images. More... | |
| float | free_memory_percentage |
| Percentage of system memory that the svl tries to leave free. More... | |
| int | num_threads |
| Maximum number of OpenMP threads that SVL can use. More... | |
| int | reproducibility |
| Enable the reproducibility of the training. More... | |
| sb_t_devices_par | devices |
| Devices used for SVL/training. More... | |
| sb_t_svl_sl_par | sl |
| Shallow Learning SVL parameters. More... | |
| sb_t_svl_dl_par | dl |
| Deep Learning SVL parameters. More... | |
| sb_t_image_circularity_type | image_circularity_type |
| Image circularity. More... | |
SVL parameters.
| sb_t_devices_par sb_t_svl_par::devices |
Devices used for SVL/training.
Used by the function sb_svl_run in case of Deep Cortex and Deep Surface projects. Currently only a single device is admitted.
| sb_t_svl_dl_par sb_t_svl_par::dl |
Deep Learning SVL parameters.
Training parameters used only by Deep Cortex and Deep Surface projects.
| sb_fp_svl_command sb_t_svl_par::fp_command |
Callback called by SVL to allow the user to decide how to continue when particular situations happen, see sb_t_svl_stop_reason .
| sb_fp_svl_pre_elaboration sb_t_svl_par::fp_pre_elaboration |
Callback to execute an image pre elaboration.
| sb_fp_svl_progress sb_t_svl_par::fp_progress |
The SVL calls this callback to notify the user the results of SVL.
| float sb_t_svl_par::free_memory_percentage |
| sb_t_image_circularity_type sb_t_svl_par::image_circularity_type |
Image circularity.
When there are objects or surfaces to be analyzed that have a circular symmetry it could be more efficient, instead of processing the image as it is, to extract the circular crown and stretch it or unrolled it so as to work in Cartesian rather than polar coordinates.
For example, if we think of extracting the circular crown from the acquired image shown in the figure starting from 12 o'clock in a clockwise direction we obtain the unrolled image shown below.
If you have to perform an ocr it is more efficient to use the unrolled circular crown instead of the original image.
In this case it is necessary to set the parameter to SB_IMAGE_CIRCULARITY_TYPE_HORIZONTAL so that the sb_svl_run function or the sb_project_detection function can adequately handle the images.
This means that the functions autonomously manage the occurrences of objects or defects that are broken between the beginning and the end of the image.
For example in the image below the letter G of the word 'image' is broken and one part is at the beginning of the image and the other at the end.
If the library is told to consider the image as circular then it is the Retina module that takes care of joining the two parts of the object in order to identify the letter 'G'.
In the results of the sb_project_detection function the barycenter of the defect or center of the sample will be placed at the beginning or at the end of the image depending on where the most bounding box of the sample or blob is located.
Circular images are managed by Retina, Deep Cortex and Deep Surface modules and by the functions sb_svl_run, sb_project_detection and sb_blobs_detection.
| char sb_t_svl_par::image_ext[64] |
Extensions of the images.
List of the extensions of the images that the function sb_svl_run searchs in the project_path.
See SB_IMAGE_FILE_EXTENSIONS for a complete list of the possible extensions.
| int sb_t_svl_par::num_threads |
Maximum number of OpenMP threads that SVL can use.
The value 0 means: half of the available threads for physical machines and all available threads for virtual machines.
| int sb_t_svl_par::reproducibility |
Enable the reproducibility of the training.
If enabled, training results are repeatable for all the SVL runs, provided that sb_t_svl_par and the dataset are the same.
Otherwise, some random initializations are applied during SVL, leading each time to different training results.
The possible values are:
| sb_t_svl_sl_par sb_t_svl_par::sl |
Shallow Learning SVL parameters.
Training parameters used only by Deep Cortex and Surface projects.
| void* sb_t_svl_par::user_data |
Pointer to data which is passed to the callbacks.
1.9.3