Update app.py
Browse files
app.py
CHANGED
@@ -8,7 +8,6 @@ import streamlit as st
|
|
8 |
from streamlit_webrtc import WebRtcMode, webrtc_streamer
|
9 |
from sample_utils.turn import get_ice_servers
|
10 |
import mediapipe as mp
|
11 |
-
from cvzone.SelfiSegmentationModule import SelfiSegmentation
|
12 |
import os
|
13 |
import time
|
14 |
|
@@ -21,11 +20,10 @@ st.title("Interactive Virtual Keyboard")
|
|
21 |
st.subheader('''Turn on the webcam and use hand gestures to interact with the virtual keyboard.
|
22 |
Use 'a' and 'd' from the keyboard to change the background.''')
|
23 |
|
24 |
-
# Initialize MediaPipe
|
25 |
mp_hands = mp.solutions.hands
|
26 |
-
hands = mp_hands.Hands(
|
27 |
mp_drawing = mp.solutions.drawing_utils
|
28 |
-
segmentor = SelfiSegmentation()
|
29 |
|
30 |
# Virtual Keyboard Layout
|
31 |
keys = [["Q", "W", "E", "R", "T", "Y", "U", "I", "O", "P"],
|
@@ -45,15 +43,6 @@ class Detection(NamedTuple):
|
|
45 |
|
46 |
result_queue: "queue.Queue[List[Detection]]" = queue.Queue()
|
47 |
|
48 |
-
# Load Background Images
|
49 |
-
listImg = os.listdir('model/street') if os.path.exists('model/street') else []
|
50 |
-
if not listImg:
|
51 |
-
st.error("Error: 'street' directory is missing or empty. Please add background images.")
|
52 |
-
st.stop()
|
53 |
-
else:
|
54 |
-
imgList = [cv2.imread(f'model/street/{imgPath}') for imgPath in listImg]
|
55 |
-
imgList = [img for img in imgList if img is not None]
|
56 |
-
|
57 |
indexImg = 0
|
58 |
output_text = ""
|
59 |
prev_key_time = [time.time()] * 2
|
@@ -61,30 +50,32 @@ prev_key_time = [time.time()] * 2
|
|
61 |
if "output_text" not in st.session_state:
|
62 |
st.session_state["output_text"] = ""
|
63 |
|
64 |
-
# Video Frame Callback with
|
65 |
def video_frame_callback(frame: av.VideoFrame) -> av.VideoFrame:
|
66 |
global indexImg, output_text
|
67 |
|
68 |
img = frame.to_ndarray(format="bgr24")
|
69 |
-
imgOut = segmentor.removeBG(img, imgList[indexImg])
|
70 |
|
71 |
# Process frame using MediaPipe
|
72 |
-
result = hands.process(cv2.cvtColor(
|
73 |
|
74 |
-
|
|
|
|
|
|
|
75 |
|
76 |
detections = []
|
77 |
if result.multi_hand_landmarks:
|
78 |
for hand_landmarks in result.multi_hand_landmarks:
|
79 |
# Draw hand landmarks
|
80 |
mp_drawing.draw_landmarks(
|
81 |
-
|
82 |
mp_drawing.DrawingSpec(color=(0, 255, 0), thickness=2, circle_radius=4),
|
83 |
mp_drawing.DrawingSpec(color=(0, 0, 255), thickness=2)
|
84 |
)
|
85 |
|
86 |
# Extract bounding box for each hand
|
87 |
-
h, w, _ =
|
88 |
x_min, y_min = w, h
|
89 |
x_max, y_max = 0, 0
|
90 |
for lm in hand_landmarks.landmark:
|
@@ -101,7 +92,7 @@ def video_frame_callback(frame: av.VideoFrame) -> av.VideoFrame:
|
|
101 |
x8, y8 = int(hand_landmarks.landmark[mp_hands.HandLandmark.INDEX_FINGER_TIP].x * w), \
|
102 |
int(hand_landmarks.landmark[mp_hands.HandLandmark.INDEX_FINGER_TIP].y * h)
|
103 |
|
104 |
-
# Distance
|
105 |
distance = np.sqrt((x8 - x4) ** 2 + (y8 - y4) ** 2)
|
106 |
click_threshold = 50
|
107 |
|
@@ -109,10 +100,10 @@ def video_frame_callback(frame: av.VideoFrame) -> av.VideoFrame:
|
|
109 |
x, y = button.pos
|
110 |
w, h = button.size
|
111 |
if x < x8 < x + w and y < y8 < y + h:
|
112 |
-
cv2.rectangle(
|
113 |
-
cv2.putText(
|
114 |
|
115 |
-
# Simulate
|
116 |
if (distance / np.sqrt(bbox[2] ** 2 + bbox[3] ** 2)) * 100 < click_threshold:
|
117 |
if time.time() - prev_key_time[0] > 2:
|
118 |
prev_key_time[0] = time.time()
|
@@ -125,7 +116,7 @@ def video_frame_callback(frame: av.VideoFrame) -> av.VideoFrame:
|
|
125 |
|
126 |
result_queue.put(detections)
|
127 |
st.session_state["output_text"] = output_text
|
128 |
-
return av.VideoFrame.from_ndarray(
|
129 |
|
130 |
# WebRTC Streamer
|
131 |
webrtc_streamer(
|
|
|
8 |
from streamlit_webrtc import WebRtcMode, webrtc_streamer
|
9 |
from sample_utils.turn import get_ice_servers
|
10 |
import mediapipe as mp
|
|
|
11 |
import os
|
12 |
import time
|
13 |
|
|
|
20 |
st.subheader('''Turn on the webcam and use hand gestures to interact with the virtual keyboard.
|
21 |
Use 'a' and 'd' from the keyboard to change the background.''')
|
22 |
|
23 |
+
# Initialize MediaPipe Hand Detection
|
24 |
mp_hands = mp.solutions.hands
|
25 |
+
hands = mp_hands.Hands(max_num_hands=1, min_detection_confidence=0.5)
|
26 |
mp_drawing = mp.solutions.drawing_utils
|
|
|
27 |
|
28 |
# Virtual Keyboard Layout
|
29 |
keys = [["Q", "W", "E", "R", "T", "Y", "U", "I", "O", "P"],
|
|
|
43 |
|
44 |
result_queue: "queue.Queue[List[Detection]]" = queue.Queue()
|
45 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
46 |
indexImg = 0
|
47 |
output_text = ""
|
48 |
prev_key_time = [time.time()] * 2
|
|
|
50 |
if "output_text" not in st.session_state:
|
51 |
st.session_state["output_text"] = ""
|
52 |
|
53 |
+
# Video Frame Callback with Logic Correction
|
54 |
def video_frame_callback(frame: av.VideoFrame) -> av.VideoFrame:
|
55 |
global indexImg, output_text
|
56 |
|
57 |
img = frame.to_ndarray(format="bgr24")
|
|
|
58 |
|
59 |
# Process frame using MediaPipe
|
60 |
+
result = hands.process(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
|
61 |
|
62 |
+
# Create Buttons
|
63 |
+
buttonList = [Button([30 + col * 105, 30 + row * 120], key)
|
64 |
+
for row, line in enumerate(keys)
|
65 |
+
for col, key in enumerate(line)]
|
66 |
|
67 |
detections = []
|
68 |
if result.multi_hand_landmarks:
|
69 |
for hand_landmarks in result.multi_hand_landmarks:
|
70 |
# Draw hand landmarks
|
71 |
mp_drawing.draw_landmarks(
|
72 |
+
img, hand_landmarks, mp_hands.HAND_CONNECTIONS,
|
73 |
mp_drawing.DrawingSpec(color=(0, 255, 0), thickness=2, circle_radius=4),
|
74 |
mp_drawing.DrawingSpec(color=(0, 0, 255), thickness=2)
|
75 |
)
|
76 |
|
77 |
# Extract bounding box for each hand
|
78 |
+
h, w, _ = img.shape
|
79 |
x_min, y_min = w, h
|
80 |
x_max, y_max = 0, 0
|
81 |
for lm in hand_landmarks.landmark:
|
|
|
92 |
x8, y8 = int(hand_landmarks.landmark[mp_hands.HandLandmark.INDEX_FINGER_TIP].x * w), \
|
93 |
int(hand_landmarks.landmark[mp_hands.HandLandmark.INDEX_FINGER_TIP].y * h)
|
94 |
|
95 |
+
# Calculate Distance and Detect Button Click
|
96 |
distance = np.sqrt((x8 - x4) ** 2 + (y8 - y4) ** 2)
|
97 |
click_threshold = 50
|
98 |
|
|
|
100 |
x, y = button.pos
|
101 |
w, h = button.size
|
102 |
if x < x8 < x + w and y < y8 < y + h:
|
103 |
+
cv2.rectangle(img, button.pos, (x + w, y + h), (0, 255, 160), -1)
|
104 |
+
cv2.putText(img, button.text, (x + 20, y + 70), cv2.FONT_HERSHEY_PLAIN, 5, (255, 255, 255), 3)
|
105 |
|
106 |
+
# Simulate Key Press if Finger Close Enough
|
107 |
if (distance / np.sqrt(bbox[2] ** 2 + bbox[3] ** 2)) * 100 < click_threshold:
|
108 |
if time.time() - prev_key_time[0] > 2:
|
109 |
prev_key_time[0] = time.time()
|
|
|
116 |
|
117 |
result_queue.put(detections)
|
118 |
st.session_state["output_text"] = output_text
|
119 |
+
return av.VideoFrame.from_ndarray(img, format="bgr24")
|
120 |
|
121 |
# WebRTC Streamer
|
122 |
webrtc_streamer(
|