This demo uses OpenCV and Microsoft Face API to detect, classify faces in an image, and possibily to recognize the persons. The puzzle is assembled live... stay tuned!
- https://docs.microsoft.com/en-us/azure/cognitive-services/face/face-api-how-to-topics/howtoidentifyfacesinimage
- https://westus.dev.cognitive.microsoft.com/docs/services/563879b61984550e40cbbe8d/operations/563879b61984550f30395236
 |
 |
|---|
I had to play a bit with OpenCV to track the face in a live video feed so that it looks a bit more interactive, but the Microsoft Face API should not be called at every frame simply for a matter of cost.
Various tests:
- OpenCV face (and eyes detection) tracking from your laptop's camera stream:
python ./test_openCV/face_eyes_detection/OpenCV2_Face_Detection.py-
same code, but for some reason in a notebook, the video capture falls back to Lena's picture with random noise changing over time: open this notebook
./test_openCV/face_eyes_detection/OpenCV2 Face Detection Lena+noise.ipynb -
Strangely this little piece of OpenCV code works fine opening your laptop's camera feed in a notebook:
import numpy as np
import cv2
cap = cv2.VideoCapture(0)
#set the width and height, and UNSUCCESSFULLY set the exposure time
cap.set(3,1080)
cap.set(4,1024)
cap.set(15, 0.1)
while True:
ret, img = cap.read()
cv2.imshow("input",img)
key = cv2.waitKey(10)
if key == 27:
break
cv2.destroyAllWindows()
cv2.VideoCapture(0).release()- OK, now this one succeeds to detect and track a face from within the notebook:
open this notebook ./test_openCV/face_eyes_detection/OpenCV_Face_Tracking_in_Notebook.ipynb