Retrain KNN classifier model in face_recognition_knn.py

• face_recognition version: latest version
• Python version: 3.6
• Operating System: Mac os High Sierra

Description

1. I want to retrain the already classified knn model.

I am saving the knn pretained model on 10,000 images. Now, i am geting 10 more new images. I dont want to train the whole model again.

Can we train the new data seperately and add to the saved model ?
(or)
Is there any parameter for scikit knn algorithm that allows us to retrain the existing model ?

I am not sure if any of the 2 ways are correct. Any ideas or references to articles on how to implement any of this is much appreciated.

2. I want to print the names of 2 nearest neighbours. what i did is:

    faces_encodings = face_recognition.face_encodings(X_img, known_face_locations=X_faces_loc)

    closest_distances = knn_clf.kneighbors(faces_encodings, n_neighbors=2)

    is_recognized = [closest_distances[0][i] <= DIST_THRESH for i in range(len(X_faces_loc))]

    print(is_recognized) # This prints [True True]

    Now i dont understand how to parse is_recognized to print the recognized values. I want to print 
    names of all is_recognized

    return [(pred) if rec for pred, rec in zip(knn_clf.predict(faces_encodings), is_recognized)] 

Any guidance on printing the neighbouring names ?

This maybe the repost of ongoing talk at https://github.com/ageitgey/face_recognition/issues/319. But that thread became too long to address. So, just reframed it here.

Thank you,

Start with face_recognition_knn.py and trace how the saved KNN model is trained, queried with knn_clf.kneighbors, and used by predict. Review the referenced issue #319 and the scikit-learn KNN behavior. Done should clarify or support adding new training data without retraining everything and provide the names of the two nearest recognized neighbors.