 |
SqueezeBrains SDK 1.18
|
|
SqueezeBrains SDK 1.18
|
Defines the results of SVL. More...
#include <sb.h>
Data Fields | |
| sb_t_par | par |
| Project parameters. | |
| sb_t_svl_stop_reason | stop_reason |
| It describes the reason why the SVL is finished. | |
| sb_t_svl_command | last_command |
| Command executed by SVL after a stop that is detailed with the parameter stop_reason. | |
| char | running_step [256] |
| Description of the current step of the training. More... | |
| sb_t_svl_err | err |
| Error details. More... | |
| sb_t_svl_res_models | models |
| Results for each model. | |
| sb_t_svl_res_level | global |
| Cumulative results of all levels of all models. | |
| sb_t_samples * | list_tp |
| List of True Positive samples. More... | |
| sb_t_samples * | list_tn |
| List of True Negative samples. More... | |
| sb_t_samples * | list_fp |
| List of False Positive samples. More... | |
| sb_t_samples * | list_fn |
| List of False Negative samples. More... | |
| sb_t_samples * | list_op |
| List of Optional Positive samples. More... | |
| sb_t_samples * | list_on |
| List of Optional Negative samples. More... | |
| sb_t_samples * | list_out_of_roi |
| List of samples out of the ROI. More... | |
| sb_t_svl_res_images * | list_img_elab |
| Results for each images. More... | |
| int | num_img_svl |
| Number of svl images of the project. More... | |
| int | num_img_test |
| Number of test images of the project. More... | |
| long long | time_ms |
| Execution time, in ms, of SVL. | |
| long long | time_ms_last_improvement |
| Time, in ms, when there was the last improvement. More... | |
| long long | time_left_ms |
| Estimated time, in ms, left to the end of the SVL. More... | |
| int | reset_done |
| The SVL reset has been done or no training has been done yet. | |
| sb_t_devices_par | devices |
| The devices used to compute the current SVL results. | |
Defines the results of SVL.
In the case of Surface, the lists of samples (list_tp, list_tn, list_fp, list_fn) have been limited to 200 elements. The 100 elements for each list are sorted by weight with the precedence of the samples used for training compared to the other samples. Sorting of samples by weight:
| sb_t_svl_err sb_t_svl_res::err |
| sb_t_samples* sb_t_svl_res::list_fn |
| sb_t_samples* sb_t_svl_res::list_fp |
| sb_t_svl_res_images* sb_t_svl_res::list_img_elab |
Results for each images.
The list exists only after sb_svl_run has been called and not after sb_project_load.
Used only by Retina and Surface projects.
| sb_t_samples* sb_t_svl_res::list_on |
| sb_t_samples* sb_t_svl_res::list_op |
| sb_t_samples* sb_t_svl_res::list_out_of_roi |
| sb_t_samples* sb_t_svl_res::list_tn |
| sb_t_samples* sb_t_svl_res::list_tp |
| int sb_t_svl_res::num_img_svl |
Number of svl images of the project.
With Deep Surface, Deep Cortex and Deep Retina projects the counter considers all the images of the training set including the validation ones and those without analysis ROI.
| int sb_t_svl_res::num_img_test |
Number of test images of the project.
With Deep Surface, Deep Cortex and Deep Retina projects the counter only considers images of the test set excluding the validation images and images without analysis ROI.
| char sb_t_svl_res::running_step[256] |
Description of the current step of the training.
The description of the training steps depends on the project type.
WithRetina projects the training steps are the following:
With Surface projects the training steps are the following:
With Deep Cortex and Deep Surface projects the training steps are the following:
| long long sb_t_svl_res::time_left_ms |
Estimated time, in ms, left to the end of the SVL.
Used only by Deep Cortex and Deep Surface projects.
| long long sb_t_svl_res::time_ms_last_improvement |
Time, in ms, when there was the last improvement.
Used for Deep Cortex and Deep Surface projects only if sb_t_svl_dl_par::save_best is enabled.
1.9.3