First of all download model from here <https://www.dropbox.com/s/tj96fsm6t6rq8ye/model-r100-arcface-ms1m-refine-v2.zip?dl=0> and put the model-r100-ii directory on './Boxes/'
install face-detection with "pip install face-detection"
your faceset csv should have this format ['id of path','xmin','ymin','xmax','ymax','id of face']
put the image paths csv and subjects csv in same folder with faceset.csv
consider the name of these 2 csv should be imagespath.csv and subjects.csv
imagespath.csv should have this format ['id of path' , 'path', 'get embeding from this path ?(flag) ']
and subjects.csv -> ['id of name ' , 'name']
after run the app :
you can click on "from csv" and choose your faceset csv (witch should have defined format)
a embs.npy will be made from that csv in long period that have the embedings dictionary of that csv . so be patient :)
please dont change imagespath and subjects manually carelessly.
load image , edit if it needs and press save button (check under button , that should be on csv mode ,it changes with click )
you can set the Auto save mode in view dropdown and just click on next button ( or push 'D') but if you dont edit labels those shape will be saved with unknown label and -1 for id
if you edit some image and save it again , it will be updated on faceset.csv
you can change constans on constants widget .
when founded 3 faces with more than second threshold it would be stop searching for similar faces .