New answer
Naster17 opened this issue · comments
What do you recommend if the photo is not of very good quality, but is taken by combining frames from the video and I leave the option of selecting faces to the neural network (to distinguish where there is a face and where there is not)
Please advise if I want to index a very large number of persons
And what do you think about:
normalization (string): Normalize the input image before feeding it to the model.
Options: base, raw, Facenet, Facenet2018, VGGFace, VGGFace2, ArcFace (default is base).
for you comb. you say me before
1- set enforce_detection argument to False if you are working with low quality images - https://github.com/serengil/deepface/blob/master/deepface/DeepFace.py#L70
2- you should use ann algorithm to index large scale database: https://sefiks.com/tag/nearest-neighbor/
3- use the default option for normalization.