Iris Tracking - Testing Raspberry Pi Capacitive TFT Screen

import cv2 as cv2
import  mediapipe as mp
import numpy as np

1.	
# Import visual functions and styles
2.	
3.	mp_drawing=mp.solutions.drawing_utils
4.	
5.	#mp_drawing_styles=mp.solutions.drawing_styles
6.	
7.	# Import 3D face landmarks detection model 
8.	
9.	mp_face_mesh=mp.solutions.face_mesh
10.	
11.	drawing_spec = mp_drawing.DrawingSpec(thickness=1, circle_radius=1)
12.	
13.	cap = cv2.VideoCapture(0)
14.	
15.	with mp_face_mesh.FaceMesh(
16.	
17.	    max_num_faces=4,
18.	
19.	    min_detection_confidence=0.5,
20.	
21.	    min_tracking_confidence=0.5) as face_mesh:
22.	
23.	  while cap.isOpened():
24.	
25.	    success, frame = cap.read()
26.	
27.	    if not success:
28.	
29.	      print("Ignoring empty camera frame.")
30.	
31.	      # If loading a video, use 'break' instead of 'continue'.
32.	
33.	      break
34.	
35.	
36.	
37.	    h, w, c = frame.shape
38.	
39.	    # image = cv2.resize(frame, (w //2, h//2))
40.	
41.	    # frame = cv2.flip(frame, 1)
42.	
43.	    image = np.copy(frame)
44.	
45.	    h2, w2, c2 = image.shape
46.	
47.	    # To improve performance, optionally mark the image as not writeable to
48.	
49.	    # pass by reference.
50.	
51.	    image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
52.	
53.	    results = face_mesh.process(image)
54.	
55.	
56.	
57.	    # Draw the face mesh annotations on the image.
58.	
59.	    image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
60.	
61.	    left_eyes = []
62.	
63.	    right_eyes = []
64.	
65.	    if results.multi_face_landmarks:
66.	
67.	      for face_landmarks in results.multi_face_landmarks:
68.	
69.	        for idx, landmark in enumerate(face_landmarks.landmark):
70.	
71.	          
72.	
73.	          if idx == 246 or idx == 159 or idx == 158  or idx == 145 or idx == 153 or idx == 190:  # left
74.	
75.	          #if idx == 53 or idx == 111:  # left
76.	
77.	            x1 = np.int(landmark.x * w2)
78.	
79.	            y1 = np.int(landmark.y * h2)
80.	
81.	            left_eyes.append((x1, y1))
82.	
83.	            #cv2.circle(image, (x1, y1), 4, (255, 0, 255), 4, cv2.LINE_AA)
84.	
85.	          if idx ==374 or idx == 380  or idx == 385 or idx == 386 or idx == 390  or idx == 414: # right
86.	
87.	          #if idx == 276 or idx == 340: # right
88.	
89.	            x1 = np.int(landmark.x * w2)
90.	
91.	            y1 = np.int(landmark.y * h2)
92.	
93.	            right_eyes.append((x1, y1))
94.	
95.	            #cv2.circle(image, (x1, y1), 4, (0, 255, 255), 4, cv2.LINE_AA)
96.	
97.	        if len(right_eyes)+len(left_eyes)==12: 
98.	
99.	         # Bounding Rectangle, wrap the found shape with a smallest rectangle. 
100.	
101.	         right_box = cv2.boundingRect(np.asarray(right_eyes))
102.	
103.	        
104.	
105.	         left_box = cv2.boundingRect(np.asarray(left_eyes))
106.	
107.	         
108.	
109.	         detect_iris(image, right_box, left_box)
110.	
111.	    cv2.imshow('MediaPipe Face Mesh', image)
112.	
113.	    if cv2.waitKey(5) & 0xFF == 27:
114.	
115.	      cv2.imwrite("D:/iris_detect_result.png", image)
116.	
117.	      break
118.	
119.	cap.release()
120.	
121.	cv2.waitKey(0)
122.	
123.	cv2.destroyAllWindows()
124.	

def detect_iris(image, right_box, left_box):
2.	
3.	   
4.	
5.	  left_roi = image[left_box[1]:left_box[1] + left_box[3], left_box[0]:left_box[0] + left_box[2]]
6.	
7.	  cv2.imshow('left_eye', left_roi)
8.	
9.	  lh, lw, lc = left_roi.shape
10.	
11.	  right_roi = image[right_box[1]:right_box[1]+right_box[3],right_box[0]:right_box[0]+right_box[2]]
12.	
13.	  cv2.imshow('right_eye', right_roi)
14.	
15.	  rh, rw, rc = right_roi.shape
16.	
17.	  if rh>0 and lh>0:
18.	
19.	    rows, cols, _ =right_roi.shape   # Save video size for backup 
20.	
21.	    gray_roi = cv2.cvtColor(right_roi, cv2.COLOR_BGR2GRAY)   # convert greyscale
22.	
23.	    gray_roi = cv2.GaussianBlur(gray_roi, (7, 7), 0)    # Gaussian filter once
24.	
25.	    _, threshold = cv2.threshold(gray_roi, 30, 255, cv2.THRESH_BINARY_INV)  #Binarization, adjust threshold as per need
26.	
27.	    contours, _ = cv2.findContours(threshold, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE) # Draw detection area
28.	
29.	    contours = sorted(contours, key=lambda x: cv2.contourArea(x), reverse=True)
30.	
31.	
32.	
33.	    for cnt in contours:
34.	
35.	        (x, y, w, h) = cv2.boundingRect(cnt)
36.	
37.	        cv2.circle(right_roi, (x + int(w/2), y + int(h/2)), int(h/2), (0, 0, 255), 3)
38.	
39.	        break
40.	
41.	
42.	
43.	    rows, cols, _ =left_roi.shape   # Save video size for backup
44.	
45.	    gray_roi = cv2.cvtColor(left_roi, cv2.COLOR_BGR2GRAY)   # convert greyscale
46.	
47.	    gray_roi = cv2.GaussianBlur(gray_roi, (7, 7), 0)    # Gaussian filter once
48.	
49.	    _, threshold = cv2.threshold(gray_roi, 30, 255, cv2.THRESH_BINARY_INV)  # Binarization, adjust threshold as per need
50.	
51.	
52.	    contours, _ = cv2.findContours(threshold, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE) # Draw detection area
53.	
54.	    contours = sorted(contours, key=lambda x: cv2.contourArea(x), reverse=True)
55.	
56.	
57.	
58.	    for cnt in contours:
59.	
60.	        (x, y, w, h) = cv2.boundingRect(cnt)
61.	
62.	        cv2.circle(left_roi, (x + int(w/2), y + int(h/2)), int(h/2), (0, 0, 255), 3)
63.	
64.	        break
65.