 |
SqueezeBrains SDK 1.18
|
|
SqueezeBrains SDK 1.18
|
Results of the SVL of a level per model. More...
#include <sb.h>
Data Fields | |
| float | goodness |
| Goodness of the training. More... | |
| float | accuracy |
| Accuracy. More... | |
| float | score |
| Level score. More... | |
| int | tp |
| Number of TRUE POSITIVE samples. More... | |
| int | fp |
| Number of FALSE POSITIVE samples. More... | |
| int | tn |
| Number of TRUE NEGATIVE samples. More... | |
| int | fn |
| Number of FALSE NEGATIVE samples. More... | |
| int | op |
| Number of OPTIONAL POSITIVE samples. More... | |
| int | on |
| Number of OPTIONAL NEGATIVE samples. More... | |
| int | out_of_roi |
| Number of Out Of ROI samples, both optional and required. More... | |
| int | mod_disabled |
| Number of samples with disabled model, both optional and required. More... | |
| int | reset |
| 1 if the SVL of the model has been resetted or when the user reset the SVL, 0 otherwise. More... | |
| int | num_samples_l |
| Number of samples of the training set used for learning. More... | |
| int | num_samples_t |
| Number of samples of the training set not used for learning. More... | |
| int | num_img_l |
| Number of images of the training set used for learning. More... | |
| int | num_tiles_l |
| Number of tiles from the training set used for learning. More... | |
| int | num_tiles_u |
| Number of tiles from the training set not used neither for learning and validation. More... | |
| int | num_tiles_v |
| Number of tiles from the training set used for validation. More... | |
| long long | num_pixels_roi_l |
| Number of pixels from the training set effectively used for learning. More... | |
| long long | num_pixels_roi_u |
| Number of pixels from the training set not used neither for learning and validation. More... | |
| long long | num_pixels_roi_v |
| Number of pixels from the training set effectively used for validation. More... | |
| long long | num_pixels_defect_l |
| Number of ROI defects pixels from the training set effectively used for training. More... | |
| long long | num_pixels_defect_u |
| Number of ROI defects pixels from the training set not used neither for learning and validation. More... | |
| long long | num_pixels_defect_v |
| Number of ROI defects pixels from the training set effectively used for validation. More... | |
| sb_t_svl_par_optimization_mode | optimization_mode |
| Optimization mode. More... | |
| char | classificator [32] |
| Classificator choosen by SVL. More... | |
| char | features [SB_PAR_FEATURES_NAMES_LEN] |
| List of the features choosen by SVL. More... | |
| char | features_available [SB_PAR_FEATURES_NAMES_LEN] |
| List of the features choosen by user for SVL. More... | |
| char * | warning |
| Warning in string format occurred during the training. | |
| sb_t_svl_res_epochs | epochs |
| Results of training of module based on Deep Learning. More... | |
Results of the SVL of a level per model.
For a Retina project you could write that:
While for a Surface project you could write that:
These two formulas are true only if there are no optional out_of_roi samples.
Usually, in Retina project, num_samples_t should be very small if compared to num_samples_l, on the contrary, in Surface projects, num_samples_t is very large because it also counts the background samples, i.e. True Negative samples.
| float sb_t_svl_res_level::accuracy |
Accuracy.
| char sb_t_svl_res_level::classificator[32] |
| sb_t_svl_res_epochs sb_t_svl_res_level::epochs |
Results of training of module based on Deep Learning.
Used only by Deep Surface and Deep Cortex projects.
| char sb_t_svl_res_level::features[SB_PAR_FEATURES_NAMES_LEN] |
List of the features choosen by SVL.
The features are separated by the SB_DELIMITER character.
Used only by Retina and Surface projects.
| char sb_t_svl_res_level::features_available[SB_PAR_FEATURES_NAMES_LEN] |
List of the features choosen by user for SVL.
The features are separated by the SB_DELIMITER character.
Used only by Retina and Surface projects.
| int sb_t_svl_res_level::fn |
Number of FALSE NEGATIVE samples.
| int sb_t_svl_res_level::fp |
Number of FALSE POSITIVE samples.
| float sb_t_svl_res_level::goodness |
| int sb_t_svl_res_level::mod_disabled |
Number of samples with disabled model, both optional and required.
| int sb_t_svl_res_level::num_img_l |
Number of images of the training set used for learning.
The count does not include images used for validation and images without analysis ROI in case of Deep Surface, Deep Cortex and Deep Retina projects.
| long long sb_t_svl_res_level::num_pixels_defect_l |
Number of ROI defects pixels from the training set effectively used for training.
The count includes all that ROI defects pixels which belong to tiles used for learning and are under analysis ROI. ROI defects pixels which are optional are not considered because not used by training. The number of pixels is counted on tiles at input resolution of the network.
Used only by Deep Surface projects
| long long sb_t_svl_res_level::num_pixels_defect_u |
Number of ROI defects pixels from the training set not used neither for learning and validation.
The count includes all that ROI defects pixels which belong to the tiles used for learning or validation that are not under analysis ROI. In addition the count includes also all ROI defects pixels belonging to unused tiles. ROI defects pixels that are optional (and thus not included in learning ones), not even in this are considered. The number of pixels is counted on tiles at input resolution of the network.
Used only by Deep Surface projects.
| long long sb_t_svl_res_level::num_pixels_defect_v |
Number of ROI defects pixels from the training set effectively used for validation.
The count includes all that ROI defects pixels which belong to tiles used for validation and under analysis ROI. ROI defects pixels which are optional are not considered because not used by validation. The number of pixels is counted on tiles at input resolution of the network.
Used only by Deep Surface projects.
| long long sb_t_svl_res_level::num_pixels_roi_l |
Number of pixels from the training set effectively used for learning.
The count includes all that pixels which belong to tiles used for learning and under analysis ROI. The number of pixels is counted on tiles at input resolution of the network.
Used only by Deep Surface projects.
| long long sb_t_svl_res_level::num_pixels_roi_u |
Number of pixels from the training set not used neither for learning and validation.
The count includes all that pixels which belong to the tiles used for learning or validation that are not under analysis ROI. In addition the count includes also all pixels belonging to unused tiles. The number of pixels is counted on tiles at input resolution of the network.
Used only by Deep Surface projects.
| long long sb_t_svl_res_level::num_pixels_roi_v |
Number of pixels from the training set effectively used for validation.
The count includes all that pixels which belong to tiles used for validation and under analysis ROI. The number of pixels is counted on tiles at input resolution of the network.
Used only by Deep Surface projects.
| int sb_t_svl_res_level::num_samples_l |
| int sb_t_svl_res_level::num_samples_t |
| int sb_t_svl_res_level::num_tiles_l |
Number of tiles from the training set used for learning.
A tile will be used for learning if a part of it is under analysis ROI.
Used only by Deep Surface projects.
| int sb_t_svl_res_level::num_tiles_u |
Number of tiles from the training set not used neither for learning and validation.
A tile will not be used for learning or validation if no part of it is under analysis ROI.
Used only by Deep Surface projects.
| int sb_t_svl_res_level::num_tiles_v |
Number of tiles from the training set used for validation.
A tile will be used for validation if a part of it is under analysis ROI. A validation tile is randomly selected according to the validaion percentage parameter.
Used only by Deep Surface projects.
| int sb_t_svl_res_level::on |
Number of OPTIONAL NEGATIVE samples.
| int sb_t_svl_res_level::op |
Number of OPTIONAL POSITIVE samples.
| sb_t_svl_par_optimization_mode sb_t_svl_res_level::optimization_mode |
| int sb_t_svl_res_level::out_of_roi |
Number of Out Of ROI samples, both optional and required.
| int sb_t_svl_res_level::reset |
1 if the SVL of the model has been resetted or when the user reset the SVL, 0 otherwise.
The value is valid only after sb_svl_run has been called and not after sb_project_load.
| float sb_t_svl_res_level::score |
| int sb_t_svl_res_level::tn |
Number of TRUE NEGATIVE samples.
| int sb_t_svl_res_level::tp |
Number of TRUE POSITIVE samples.
1.9.3