Convolutional Neural Networks On Ninapro datasets
-. usefulFcns.py
1. SplitOnColumn. split a 2D numpy array on its Column
to a 3D numpy array
2. BuildNewlyDir.
newly build an empty directory.
if it exists, delete it, and then build a new one.
or, build a new one.
-. FeaturesExtraction.py
many features extraction functions sets
and an extractSlidingWindow methods to extract [feStr] with sliding window of [LI-LW]
-. DataPreparation.py
1. getRawDict from [.mat] file
2. getRmsImagesLabels from rawDict, to construct two generalized numpy arrays 3D[Images]mx16x30 and 2D[Labels]mx8
-. scriptDataPreparation.py a work-through script
1. read from .mat file
2. extract [RMS] feature
3. write 3D[Images]mx16x30 and 2D[Labels]mx8 to file 'ninaRmsImagesLabels.pkl' with pickle module
4. read 3D[Images]mx16x30 and 2D[Labels]mx8 from file 'ninaRmsImagesLabels.pkl' with pickle module
5. write == read ? checking
6. time duration computation for every part.
-. classNinapro.py
1. read [Images]&[Labels] from .pkl file
2. split [Images]&[Labels] to Train-Test-Validate parts with a proportion
3. next_batch for usage during CNN training.