notes.txt

SVM accuracy for test scores increases to 95% if the letters q,x,g,h are removed
from the set dataset_1. Other letters seem to be accurate enough.

Helper folder in ML is to have the models as an importable package
Just import them in other script and pass the dataframe to find_accuracy() Function
to get an object containing train and test accuracies

Dataset 4 has palmNormal feature included

Palm Normal is now changed to palm direction...manually calculated during feature extraction
0 - Not found
1 - Down
2 - Up
3 - Left
4 - Right
5 - Front
6 - Back

SVM only :
Accuracy of set HKUV increases from 98.1 to 98.9 if only included
my_list = ['thumb_index','index_middle','ring_pinky','thumb_center_distance',
'index_center_distance','middle_center_distance','ring_center_distance','pinky_center_distance',
'palm_direction','index_direction_x','index_direction_y','index_direction_z','middle_direction_x',
'middle_direction_y','middle_direction_z','label']


Accuracy of set DRP increases from 98.1 to 98.5 if only included
['index_meta_proxi','index_proxi_inter',
'middle_meta_proxi','middle_proxi_inter','thumb_index','index_middle','index_direction_x',
'index_direction_y','index_direction_z','middle_direction_x','middle_direction_y','middle_direction_z'
,'label']