Problems with face finding

• face_recognition version: any
• Python version: 3.5
• Operating System: raspbian stretch (so I am using raspberry pi3 b+)
• Camera: PS3 Eye

Description

Can't find face in any picture if half of face is more darker than another. How can I set sensetivity of face finding?

What I Did

Here is code I use

import face_recognition
import cv2

#...
#Generating face encodings...
#...

vc = cv2.VideoCapture(0)
user_counter = 0
is_capturing, frame = vc.read()
cv2.imwrite('img.jpg', frame)
face_locations = face_recognition.face_locations(frame)
face_encodings = face_recognition.face_encodings(frame, face_locations)

if len(face_encodings) > 0:
    user_counter += 1
    
if user_counter > 0:
    print('hello')
    user = face_detection(face_encodings)
    if user in known_faces:
        print('welcome, ' + user)
    else:
        print('who is there?')
    user_counter = 0
else:
    print("no faces")

So I always get "no faces" in output. To work normaly I need an ideal light, it 's not always possible.
@ageitgey can you nelp me?

Start with the provided Python snippet and the face_recognition.face_locations and face_recognition.face_encodings entry points. Reproduce the no-face result using the stated Raspberry Pi, PS3 Eye camera, and lighting conditions, then compare behavior under different lighting. Done means the sensitivity behavior and a concrete resolution are established.