Skip to content

Instantly share code, notes, and snippets.

@Asadullah-Dal17

Asadullah-Dal17/mediapipe_landmarks_detection.py

Last active September 1, 2021 07:03
Show Gist options

  • Select an option

    Learn more about clone URLs
  • Save Asadullah-Dal17/fd71c31bac74ee84e6a31af50fa62961 to your computer and use it in GitHub Desktop.

Select an option

Learn more about clone URLs
Save Asadullah-Dal17/fd71c31bac74ee84e6a31af50fa62961 to your computer and use it in GitHub Desktop.
Download ZIP
Mediapipe landmarks detection for each individual part, like an face-oval, eyes, lips, and Eyebrows
Raw
mediapipe_landmarks_detection.py
'''
Author: Asadullah Dal
Youtube Channel: https://www.youtube.com/c/aiphile
'''
import cv2 as cv
import numpy as np
import mediapipe as mp
import utils
import time
# Landmarks indices for different parts of face, mediapipe.
# face bounder indices
FACE_OVAL=[ 10, 338, 297, 332, 284, 251, 389, 356, 454, 323, 361, 288, 397, 365, 379, 378, 400, 377, 152, 148, 176, 149, 150, 136, 172, 58, 132, 93, 234, 127, 162, 21, 54, 103,67, 109]
# lips indices for Landmarks
LIPS=[ 61, 146, 91, 181, 181, 84, 17, 314, 405, 321, 375, 61, 185, 40, 39, 37,0 ,267 ,269 ,270 ,409, 78, 95 ,88 ,178 ,87, 14, 317, 402, 318, 324, 78, 191, 80, 81, 82, 13, 312, 311, 310, 415 ]
# Left eyes indices
LEFT_EYE =[ 362, 382, 381, 380, 374, 373, 390, 249, 263, 466, 388, 387, 386, 385,384, 398 ]
LEFT_EYEBROW =[ 336, 296, 334, 293, 300, 276, 283, 282, 295, 285 ]
# right eyes indices
RIGHT_EYE=[ 33, 7, 163, 144, 145, 153, 154, 155, 133, 173, 157, 158, 159, 160, 161 , 246 ]
RIGHT_EYEBROW=[ 70, 63, 105, 66, 107, 55, 65, 52, 53, 46 ]
# Variables
# Frame per seconds
frame_counter =0
# Setting up mediapipe
map_face_mesh= mp.solutions.face_mesh
# Different function
def faceLandmarksDetector(img, result, draw=False):
# image width and height
img_height, img_width = img.shape[:2]
# getting all the landmark normalized cordinate(x,y) in the image
# multiplying these coordinate with width and height we get image coordinates
mesh_cord_point = [ (int(p.x*img_width) , int(p.y*img_height)) for p in results.multi_face_landmarks[0].landmark]
if draw :
[cv.circle(img, point_cord, 2, utils.GREEN, -1) for point_cord in mesh_cord_point]
# print()
return mesh_cord_point
# setting up camera
cap = cv.VideoCapture(0)
# configring mediapipe for face mesh detection
with map_face_mesh.FaceMesh( min_detection_confidence=0.5, min_tracking_confidence=0.5 ) as face_mesh:
# string video/webcame feed here
# initial time set here
starting_time =time.time()
while True:
ret, frame = cap.read()
frame_counter +=1
# converting color space from BGR to RGB
rgb_frame = cv.cvtColor(frame, cv.COLOR_BGR2RGB)
# getting the frame width
height, width = frame.shape[:2]
# getting the landmarks
results = face_mesh.process(rgb_frame)
# checking if landmarks are detected or not
if results.multi_face_landmarks:
# calling faceLandmarksDetector function and getting coordinate of each point in face mesh
mesh_cords =faceLandmarksDetector(img=frame, result=results)
frame = utils.fillPolyTrans(img=frame,points=[mesh_cords[p] for p in FACE_OVAL],color=utils.BLACK, opacity=0.4)
# draw lips landmarks portion
frame = utils.fillPolyTrans(img=frame,points=[mesh_cords[p] for p in LIPS],color=utils.WHITE, opacity=0.4)
frame = utils.fillPolyTrans(img=frame,points=[mesh_cords[p] for p in LEFT_EYE],color=utils.YELLOW, opacity=0.4)
frame = utils.fillPolyTrans(img=frame,points=[mesh_cords[p] for p in LEFT_EYEBROW],color=utils.GREEN, opacity=0.4)
frame = utils.fillPolyTrans(img=frame,points=[mesh_cords[p] for p in RIGHT_EYE],color=utils.YELLOW, opacity=0.4)
frame = utils.fillPolyTrans(img=frame,points=[mesh_cords[p] for p in RIGHT_EYEBROW],color=utils.GREEN, opacity=0.4)
# calculating the end time of total frames
end_time = time.time()- starting_time
# calculating the frames per seconds
fps = frame_counter/end_time
cv.imshow('frame',frame)
key = cv.waitKey(1)
if key==ord('q'):
break
cv.destroyAllWindows()
cap.release()
Raw
utils.py
'''
Author: Asadullah Dal
Youtube Channel: https://www.youtube.com/c/aiphile
'''
import cv2 as cv
import numpy as np
# colors
# values =(blue, green, red) opencv accepts BGR values not RGB
BLACK = (0,0,0)
WHITE = (255,255,255)
BLUE = (255,0,0)
RED = (0,0,255)
CYAN = (255,255,0)
YELLOW =(0,255,255)
MAGENTA = (255,0,255)
GRAY = (128,128,128)
GREEN = (0,255,0)
PURPLE = (128,0,128)
ORANGE = (0,165,255)
PINK = (147,20,255)
points_list =[(200, 300), (150, 150), (400, 200)]
def drawColor(img, colors):
x, y = 0,10
w, h = 20, 30
for color in colors:
x += w+5
# y += 10
cv.rectangle(img, (x-6, y-5 ), (x+w+5, y+h+5), (10, 50, 10), -1)
cv.rectangle(img, (x, y ), (x+w, y+h), color, -1)
def textWithBackground(img, text, font, fontScale, textPos, textThickness=1,textColor=(0,255,0), bgColor=(0,0,0), pad_x=3, pad_y=3, bgOpacity=0.5):
"""
Draws text with background, with control transparency
@param img:(mat) which you want to draw text
@param text: (string) text you want draw
@param font: fonts face, like FONT_HERSHEY_COMPLEX, FONT_HERSHEY_PLAIN etc.
@param fontScale: (double) the size of text, how big it should be.
@param textPos: tuple(x,y) position where you want to draw text
@param textThickness:(int) fonts weight, how bold it should be
@param textPos: tuple(x,y) position where you want to draw text
@param textThickness:(int) fonts weight, how bold it should be.
@param textColor: tuple(BGR), values -->0 to 255 each
@param bgColor: tuple(BGR), values -->0 to 255 each
@param pad_x: int(pixels) padding of in x direction
@param pad_y: int(pixels) 1 to 1.0 (), controls transparency of text background
@return: img(mat) with draw with background
"""
(t_w, t_h), _= cv.getTextSize(text, font, fontScale, textThickness) # getting the text size
x, y = textPos
overlay = img.copy() # coping the image
cv.rectangle(overlay, (x-pad_x, y+ pad_y), (x+t_w+pad_x, y-t_h-pad_y), bgColor,-1) # draw rectangle
new_img = cv.addWeighted(overlay, bgOpacity, img, 1 - bgOpacity, 0) # overlaying the rectangle on the image.
cv.putText(new_img,text, textPos,font, fontScale, textColor,textThickness ) # draw in text
img = new_img
return img
def textBlurBackground(img, text, font, fontScale, textPos, textThickness=1,textColor=(0,255,0),kneral=(33,33) , pad_x=3, pad_y=3):
"""
Draw text with background blured, control the blur value, with kernal(odd, odd)
@param img:(mat) which you want to draw text
@param text: (string) text you want draw
@param font: fonts face, like FONT_HERSHEY_COMPLEX, FONT_HERSHEY_PLAIN etc.
@param fontScale: (double) the size of text, how big it should be.
@param textPos: tuple(x,y) position where you want to draw text
@param textThickness:(int) fonts weight, how bold it should be.
@param textColor: tuple(BGR), values -->0 to 255 each
@param kneral: tuple(3,3) int as odd number: higher the value, more blurry background would be
@param pad_x: int(pixels) padding of in x direction
@param pad_y: int(pixels) padding of in y direction
@return: img mat, with text drawn, with background blured
call the function:
img =textBlurBackground(img, 'Blured Background Text', cv2.FONT_HERSHEY_COMPLEX, 0.9, (20, 60),2, (0,255, 0), (49,49), 13, 13 )
"""
(t_w, t_h), _= cv.getTextSize(text, font, fontScale, textThickness) # getting the text size
x, y = textPos
blur_roi = img[y-pad_y-t_h: y+pad_y, x-pad_x:x+t_w+pad_x] # croping Text Background
img[y-pad_y-t_h: y+pad_y, x-pad_x:x+t_w+pad_x]=cv.blur(blur_roi, kneral) # merging the blured background to img
cv.putText(img,text, textPos,font, fontScale, textColor,textThickness )
# cv.imshow('blur roi', blur_roi)
# cv.imshow('blured', img)
return img
def fillPolyTrans(img, points, color, opacity):
"""
@param img: (mat) input image, where shape is drawn.
@param points: list [tuples(int, int) these are the points custom shape,FillPoly
@param color: (tuples (int, int, int)
@param opacity: it is transparency of image.
@return: img(mat) image with rectangle draw.
"""
list_to_np_array = np.array(points, dtype=np.int32)
overlay = img.copy() # coping the image
cv.fillPoly(overlay,[list_to_np_array], color )
new_img = cv.addWeighted(overlay, opacity, img, 1 - opacity, 0)
# print(points_list)
img = new_img
return img
def rectTrans(img, pt1, pt2, color, thickness, opacity):
"""
@param img: (mat) input image, where shape is drawn.
@param pt1: tuple(int,int) it specifies the starting point(x,y) os rectangle
@param pt2: tuple(int,int) it nothing but width and height of rectangle
@param color: (tuples (int, int, int), it tuples of BGR values
@param thickness: it thickness of board line rectangle, if (-1) passed then rectangle will be fulled with color.
@param opacity: it is transparency of image.
@return:
"""
overlay = img.copy()
cv.rectangle(overlay, pt1, pt2, color, thickness)
new_img = cv.addWeighted(overlay, opacity, img, 1 - opacity, 0) # overlaying the rectangle on the image.
img = new_img
return img
def main():
cap = cv.VideoCapture('Girl.mp4')
counter =0
while True:
success, img = cap.read()
# img = np.zeros((1000,1000, 3), dtype=np.uint8)
img=rectTrans(img, pt1=(30, 320), pt2=(160, 260), color=(0,255,255),thickness=-1, opacity=0.6)
img =fillPolyTrans(img=img, points=points_list, color=(0,255,0), opacity=.5)
drawColor(img, [BLACK,WHITE ,BLUE,RED,CYAN,YELLOW,MAGENTA,GRAY ,GREEN,PURPLE,ORANGE,PINK])
textBlurBackground(img, 'Blured Background Text', cv.FONT_HERSHEY_COMPLEX, 0.8, (60, 140),2, YELLOW, (71,71), 13, 13)
img=textWithBackground(img, 'Colored Background Texts', cv.FONT_HERSHEY_SIMPLEX, 0.8, (60,80), textThickness=2, bgColor=GREEN, textColor=BLACK, bgOpacity=0.7, pad_x=6, pad_y=6)
imgGray = cv.cvtColor(img, cv.COLOR_BGR2GRAY)
# cv.imwrite('color_image.png', img)
counter +=1
cv.imshow('img', img)
cv.imwrite(f'image/image_{counter}.png', img)
if cv.waitKey(1) ==ord('q'):
break
if __name__ == "__main__":
main()
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment