neuropoly / spinalcordtoolbox / 8812 / 2

67%

master: 59%

push

travis-ci

Added CNN for centerline detection, brain detection and added possibility for 3d CNN kernel (#1822)

* centerline detection either SVM either CNN + argument for brain or not brain section contained in the input image

* add the nn_architecture for centerline detection

* add param t2s

* add t2s t1 dwi ctr parameters + speed in case of brain section + reduce bin threshold to avoid under segmentation

* "-brain" more explicit than "-no-brain"

* make -ctr description more explicit

* kernel choice (2d or 3d) + bin the = 0.001 if t1 like else 0.5

* add 3d architecture

* add nyul normalization + allow 3d

* integrate nyul standardization

* arrange default params

* code documentation

* clean 3d architecture

* add models on OSF

* fix variable not defined

* scripts/sct_deepseg_sc.py

* recover inference pipeline of master for segment_2d

* -ctr renamed in -centerline

* update OSF link

* add warning type to print

* threshold updated

* updated URLs for deepseg_sc models

* changed default value for centerline flag with t2s: cnn->svm

50 of 64 branches covered (78.13%)

Branch coverage included in aggregate %.

124 of 140 relevant lines covered (88.57%)

0.89 hits per line