File size: 7,064 Bytes
8d5b7a9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
import cv2
import numpy as np
from PIL import Image
import mediapipe as mp
import time

"""
This code can not be run on HuggingFace's Spaces App due to constraints 
brought by Gradio's limited input and output functionality

This features both more and less functions

- Same "pen-holding" gesture to write, let go of the pen to lift off the "paper" 
- Open palm facing front gesture to save a copy of the paper to home directory
- Thumbs up gesture to clear the page

*** Install dependencies from requirements.txt
*** packages.txt is device dependent
"""


def find_hands(brain, img):
    img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB) # opencv image is in BGR form but mp is trained with RGB
    results = brain.process(img_rgb) # process finds the hands and outputs classification and 21 landmarks for each hand
    hands_landmarks = [] # initializing array to hold the dictionary for the hands
    h, w, _ = img.shape # get height and width of image for scaling
    if results.multi_hand_landmarks:
        for hand_type, hand_lms in zip(results.multi_handedness, results.multi_hand_landmarks): # elegant solution for mp list object traversal
            hand = {} # initializing dict for each hand
            lm_list = [] # landmarks array for all 21 point of the hand
            for lm in hand_lms.landmark:
                px, py, pz = int(lm.x * w), int(lm.y * h), int(lm.z * w) # scaling landmark points to image size for frame coordinates
                lm_list.append([px, py, pz])

            hand["lm_list"] = lm_list # add "lm_list" key for all landmark points of the hand
            hand["type"] = hand_type.classification[0].label # adds the label (left/right) for the hand
            hands_landmarks.append(hand) # appends the dict
    return hands_landmarks


def is_drawing(index, thumb): # proximity function with arbitrary threshold
    npindex = np.array((index[0], index[1]))
    npthumb = np.array((thumb[0], thumb[1]))
    if np.linalg.norm(npindex - npthumb) < 30:
        return True
    else:
        return False


def save(landmarks): # brute force finger orientation checking
    if landmarks[8][1] < landmarks[6][1]:
        if landmarks[12][1] < landmarks[10][1]:
            if landmarks[16][1] < landmarks[14][1]:
                if landmarks[20][1] < landmarks[18][1]:
                    return True
    else:
        return False


def clear(landmarks): # brute force finger orientation checking
    if landmarks[4][1] < landmarks[3][1] < landmarks[2][1] < landmarks[8][1]:
        return True
    else:
        return False


DOMINANT_HAND = "Right"

width, height = 1280, 720
width_, height_, = 256, 144

drawing_flag = False
sleepy_time = time.time()


if __name__ == '__main__':
    cam = cv2.VideoCapture(0)
    cam.set(3, width)
    cam.set(4, height)

    detector = mp.solutions.hands.Hands(min_detection_confidence=0.8) # initialize mp model
    # paper = np.zeros((width, height, 4), np.uint8)
    paper = np.zeros((height, width, 3), dtype=np.uint8) # create blank page
    paper.fill(255)

    past_holder = () # coordinates holder
    palette = cv2.imread('palette.jpg')

    output_frames = []
    page_num = 0
    # runny = 1
    color = (0, 0, 0)
    while True:
        # runny -= 1
        x, rgb_image = cam.read()
        rgb_image_f = cv2.flip(np.asanyarray(rgb_image), 1)

        hands = find_hands(detector, rgb_image_f)

        try:
            if hands:
                hand1 = hands[0] if hands[0]["type"] == DOMINANT_HAND else hands[1]
                lm_list1 = hand1["lm_list"]  # List of 21 Landmarks
                handedness = hand1["type"]

                if handedness == DOMINANT_HAND:
                    idx_coords = lm_list1[8][0], lm_list1[8][1]  # 0 is width (bigger)
                    # print(idx_coords)
                    cv2.circle(rgb_image_f, idx_coords, 5, color, cv2.FILLED)

                    if idx_coords[1] < 72: # brute force but should be extremely marginally faster lol
                        if idx_coords[0] < 142:  # red
                            color = (0, 0, 255)
                        if 142 < idx_coords[0] < 285:  # orange
                            color = (0, 115, 255)
                        if 285 < idx_coords[0] < 426:  # yellow
                            color = (0, 229, 255)
                        if 426 < idx_coords[0] < 569:  # green
                            color = (0, 195, 88)
                        if 569 < idx_coords[0] < 711:  # blue
                            color = (195, 85, 0)
                        if 711 < idx_coords[0] < 853:  # indigo
                            color = (195, 0, 68)
                        if 853 < idx_coords[0] < 996:  # violet
                            color = (195, 0, 143)
                        if 996 < idx_coords[0] < 1137:  # black
                            color = (0, 0, 0)
                        if 1137 < idx_coords[0]:  # white / eraser
                            color = (255, 255, 255)

                    if len(past_holder) and drawing_flag: # start drawing
                        cv2.line(paper, past_holder, idx_coords, color, 5)
                        cv2.line(rgb_image_f, past_holder, idx_coords, color, 5)
                        # paper[idx_coords[0]][idx_coords[1]][0] = 255
                        # paper[idx_coords[0]][idx_coords[1]][3] = 255
                        cv2.circle(rgb_image_f, idx_coords, 5, color, cv2.FILLED)

                    if save(lm_list1) and time.time() - sleepy_time > 3: # save page, 3 secs arbitrary, just to not iterate every loop iteration
                        paper[0:height_, w - width_: w] = 255
                        paper = cv2.cvtColor(paper, cv2.COLOR_BGR2RGB)
                        im = Image.fromarray(paper)
                        im.save("paper%s.png" % page_num)
                        print("saved")
                        sleepy_time = time.time()
                        paper = cv2.cvtColor(paper, cv2.COLOR_RGB2BGR)
                        page_num += 1

                    if clear(lm_list1) and time.time() - sleepy_time > 3: # clear page
                        paper = np.zeros((height, width, 3), dtype=np.uint8)
                        paper.fill(255)
                        print("page cleared")
                        sleepy_time = time.time()

                    past_holder = idx_coords

            if is_drawing(idx_coords, lm_list1[4]):  # 4 is thumb for intuitive "hold pen" to draw
                drawing_flag = True
            else:
                drawing_flag = False

        except:
            pass

        finally:
            rgb_image_f[0:72, ] = palette
            presenter = cv2.resize(rgb_image_f, (width_, height_))
            h, w, _ = rgb_image_f.shape
            paper[0:height_, w - width_: w] = presenter
            cv2.imshow("Image", rgb_image_f)
            cv2.imshow("paper", paper)
            key = cv2.waitKey(1)
            if key & 0xFF == ord('q') or key == 27:  # Press esc or 'q' to close the image window
                break