digital makeup for real time

[roiksail]

hello everybody,how i use digital makeup for real time?

[Titaniumtown]

Yea I Would like to know this as well!

[navalchand]

Please find below script to do digital makeup for real time. Currently, I'm using PIL. If we can change to opencv it will be much faster. At least for landmark, it is fast compared to PIL.

For faster performance make sure to add down-scaling.

down_sampling = True

Code:

import cv2
import face_recognition
from PIL import Image, ImageDraw
import numpy as np

def PIL2array(img):
    """ Convert a PIL/Pillow image to a numpy array """
    return np.array(img.getdata(),
        np.uint8).reshape(img.size[1], img.size[0], 3)

def up_sample(landmark_list , sample_size=4):
    for face_landmark in landmark_list:
        if len(face_landmark) > 1:
            for key in face_landmark.keys():
                face_landmark[key] = [(w[0]*sample_size , w[1]*sample_size) for w in face_landmark[key]]
    return landmark_list

class FaceLandMarkDetection:

    def predict(self , frame):
        face_landmarks = face_recognition.face_landmarks(frame)
        if down_sampling:
            self.face_landmarks = up_sample(face_landmarks)
        else:
            self.face_landmarks = face_landmarks

    def plot(self , frame):
        pil_image = Image.fromarray(frame)
        for face_landmarks in self.face_landmarks:
            if len(face_landmarks) > 1:
                d = ImageDraw.Draw(pil_image, 'RGBA')

                # Make the eyebrows into a nightmare
                d.polygon(face_landmarks['left_eyebrow'], fill=(68, 54, 39, 128))
                d.polygon(face_landmarks['right_eyebrow'], fill=(68, 54, 39, 128))
                d.line(face_landmarks['left_eyebrow'], fill=(68, 54, 39, 150), width=5)
                d.line(face_landmarks['right_eyebrow'], fill=(68, 54, 39, 150), width=5)

                # Gloss the lips
                d.polygon(face_landmarks['top_lip'], fill=(150, 0, 0, 128))
                d.polygon(face_landmarks['bottom_lip'], fill=(150, 0, 0, 128))
                d.line(face_landmarks['top_lip'], fill=(150, 0, 0, 64), width=8)
                d.line(face_landmarks['bottom_lip'], fill=(150, 0, 0, 64), width=8)

                # Sparkle the eyes
                d.polygon(face_landmarks['left_eye'], fill=(255, 255, 255, 30))
                d.polygon(face_landmarks['right_eye'], fill=(255, 255, 255, 30))

                # Apply some eyeliner
                d.line(face_landmarks['left_eye'] + [face_landmarks['left_eye'][0]], fill=(0, 0, 0, 110), width=6)
                d.line(face_landmarks['right_eye'] + [face_landmarks['right_eye'][0]], fill=(0, 0, 0, 110), width=6)

        return PIL2array(pil_image)

video_capture = cv2.VideoCapture(0)
video_capture.set(cv2.CAP_PROP_FRAME_WIDTH, 1920)
video_capture.set(cv2.CAP_PROP_FRAME_HEIGHT, 1080)

face_landmark_detection = FaceLandMarkDetection()
down_sampling = True
while True:
    # Grab a single frame of video
    ret, frame = video_capture.read()

    if down_sampling:
        small_frame = cv2.resize(frame, (0, 0), fx=0.25, fy=0.25)

        # Find all facial features in all the faces in the image
        face_landmark_detection.predict(small_frame)
    else:
        # Find all facial features in all the faces in the image
        face_landmark_detection.predict(frame)

    ### plot 
    frame = face_landmark_detection.plot(frame)

    # Display the resulting image
    cv2.imshow('Video', frame)

    # Hit 'q' on the keyboard to quit!
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

# Release handle to the webcam
video_capture.release()
cv2.destroyAllWindows()

[IEWbgfnYDwHRoRRSKtkdyMDUzgdwuBYgDKtDJWd]

Glad to see others are interested in this as well, its the reason I wanted to test all this out. I managed to get it up and running, and plan to try video next thanks to naval's code.

However, I am wondering how we could expand on the photorealism. obviously there are other digital makeup libraries out there, but I would still like to build my own interpretation. From what i can see, it appears we can specify as many lines and polygons as we wish per landmark. And from there we can specify +/- widths and colors. I suppose if we were able to apply polygons with alpha transparencies, we could go further by layering colors and alpha transparencies with differing brightness and contrast. (although ideally I would like to be able to overlay pngs/vectors scaled to the landmark sizes)I am well versed in graphic manipulation but not so much python. And I know you guys know lots more than me, so interested in hearing your takes. Thanks.

[tyagi-py]

Glad to see others are interested in this as well, its the reason I wanted to test all this out. I managed to get it up and running, and plan to try video next thanks to naval's code.

However, I am wondering how we could expand on the photorealism. obviously there are other digital makeup libraries out there, but I would still like to build my own interpretation. From what i can see, it appears we can specify as many lines and polygons as we wish per landmark. And from there we can specify +/- widths and colors. I suppose if we were able to apply polygons with alpha transparencies, we could go further by layering colors and alpha transparencies with differing brightness and contrast. (although ideally I would like to be able to overlay pngs/vectors scaled to the landmark sizes)I am well versed in graphic manipulation but not so much python. And I know you guys know lots more than me, so interested in hearing your takes. Thanks.

Nice thought! Can you tell me some python and open-cv based makeup libraries ?