Update app.py
Browse files
app.py
CHANGED
@@ -1,14 +1,3 @@
|
|
1 |
-
import logging
|
2 |
-
import queue
|
3 |
-
from typing import List, NamedTuple
|
4 |
-
import av
|
5 |
-
import cv2
|
6 |
-
import numpy as np
|
7 |
-
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 |
-
import os
|
12 |
import time
|
13 |
|
14 |
# Logger Setup
|
@@ -21,41 +10,13 @@ st.subheader('''Turn on the webcam and use hand gestures to interact with the vi
|
|
21 |
|
22 |
# Initialize MediaPipe and Background Segmentor
|
23 |
mp_hands = mp.solutions.hands
|
24 |
-
hands = mp_hands.Hands(max_num_hands=1, min_detection_confidence=0.5)
|
25 |
-
mp_drawing = mp.solutions.drawing_utils
|
26 |
-
|
27 |
-
# Virtual Keyboard Layout
|
28 |
-
keys = [["Q", "W", "E", "R", "T", "Y", "U", "I", "O", "P"],
|
29 |
-
["A", "S", "D", "F", "G", "H", "J", "K", "L", ";"],
|
30 |
["Z", "X", "C", "V", "B", "N", "M", ",", ".", "/"]]
|
31 |
|
32 |
class Button:
|
33 |
-
def
|
34 |
self.pos = pos
|
35 |
self.size = size
|
36 |
self.text = text
|
37 |
-
|
38 |
-
class Detection(NamedTuple):
|
39 |
-
label: str
|
40 |
-
score: float
|
41 |
-
box: np.ndarray
|
42 |
-
|
43 |
-
result_queue: "queue.Queue[List[Detection]]" = queue.Queue()
|
44 |
-
|
45 |
-
indexImg = 0
|
46 |
-
output_text = ""
|
47 |
-
prev_key_time = [time.time()] * 2
|
48 |
-
|
49 |
-
if "output_text" not in st.session_state:
|
50 |
-
st.session_state["output_text"] = ""
|
51 |
-
|
52 |
-
# Video Frame Callback with Your Logic
|
53 |
-
def video_frame_callback(frame: av.VideoFrame) -> av.VideoFrame:
|
54 |
-
global indexImg, output_text
|
55 |
-
|
56 |
-
img = frame.to_ndarray(format="bgr24")
|
57 |
-
result = hands.process(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
|
58 |
-
|
59 |
# Create the keyboard buttons
|
60 |
buttonList = []
|
61 |
h, w = img.shape[:2]
|
@@ -76,40 +37,14 @@ def video_frame_callback(frame: av.VideoFrame) -> av.VideoFrame:
|
|
76 |
|
77 |
# Draw Keyboard Buttons
|
78 |
for button in buttonList:
|
79 |
-
x, y = button.pos
|
80 |
-
bw, bh = button.size
|
81 |
-
cv2.rectangle(img, (x, y), (x + bw, y + bh), (200, 200, 200), -1)
|
82 |
-
cv2.putText(img, button.text, (x + int(0.2 * bw), y + int(0.7 * bh)), cv2.FONT_HERSHEY_PLAIN, font_scale, (0, 0, 0), font_thickness)
|
83 |
-
|
84 |
-
detections = []
|
85 |
-
if result.multi_hand_landmarks:
|
86 |
-
for hand_landmarks in result.multi_hand_landmarks:
|
87 |
-
mp_drawing.draw_landmarks(
|
88 |
-
img, hand_landmarks, mp_hands.HAND_CONNECTIONS,
|
89 |
-
mp_drawing.DrawingSpec(color=(0, 255, 0), thickness=2, circle_radius=4),
|
90 |
-
mp_drawing.DrawingSpec(color=(0, 0, 255), thickness=2)
|
91 |
-
)
|
92 |
-
|
93 |
-
h, w, _ = img.shape
|
94 |
-
x_min, y_min = w, h
|
95 |
-
x_max, y_max = 0, 0
|
96 |
-
for lm in hand_landmarks.landmark:
|
97 |
-
x, y = int(lm.x * w), int(lm.y * h)
|
98 |
-
x_min, y_min = min(x_min, x), min(y_min, y)
|
99 |
-
x_max, y_max = max(x_max, x), max(y_max, y)
|
100 |
-
|
101 |
-
bbox = [x_min, y_min, x_max - x_min, y_max - y_min]
|
102 |
-
detections.append(Detection(label="Hand", score=0.5, box=np.array(bbox)))
|
103 |
-
|
104 |
x4, y4 = int(hand_landmarks.landmark[mp_hands.HandLandmark.THUMB_TIP].x * w), int(hand_landmarks.landmark[mp_hands.HandLandmark.THUMB_TIP].y * h)
|
105 |
x8, y8 = int(hand_landmarks.landmark[mp_hands.HandLandmark.INDEX_FINGER_TIP].x * w), int(hand_landmarks.landmark[mp_hands.HandLandmark.INDEX_FINGER_TIP].y * h)
|
106 |
|
107 |
-
distance = np.sqrt((x8 - x4)
|
108 |
-
click_threshold = 0.2*np.sqrt(bbox[2]
|
109 |
|
110 |
for button in buttonList:
|
111 |
x, y = button.pos
|
112 |
-
bw, bh = button.size
|
113 |
if x < x8 < x + bw and y < y8 < y + bh:
|
114 |
cv2.rectangle(img, (x, y), (x + bw, y + bh), (0, 255, 160), -1)
|
115 |
cv2.putText(img, button.text, (x + int(0.2 * bw), y + int(0.7 * bh)), cv2.FONT_HERSHEY_PLAIN, font_scale, (255, 255, 255), font_thickness)
|
@@ -124,11 +59,13 @@ def video_frame_callback(frame: av.VideoFrame) -> av.VideoFrame:
|
|
124 |
else:
|
125 |
output_text += ' ' # Add space
|
126 |
# Position and dimensions for the rectangle
|
|
|
127 |
text_x = int(0.05 * w)
|
128 |
text_y = int(0.70 * h)
|
129 |
text_width = int(0.9 * w) # Adjust width as needed
|
130 |
text_height = int(0.1 * h) # Adjust height as needed
|
131 |
# Draw the rectangle
|
|
|
132 |
cv2.rectangle(img,
|
133 |
(text_x, text_y - text_height), # Top-left corner
|
134 |
(text_x + text_width, text_y), # Bottom-right corner
|
@@ -144,10 +81,6 @@ def video_frame_callback(frame: av.VideoFrame) -> av.VideoFrame:
|
|
144 |
return av.VideoFrame.from_ndarray(img, format="bgr24")
|
145 |
|
146 |
# WebRTC Streamer
|
147 |
-
webrtc_streamer(
|
148 |
-
key="virtual-keyboard",
|
149 |
-
mode=WebRtcMode.SENDRECV,
|
150 |
-
rtc_configuration={"iceServers": get_ice_servers(), "iceTransportPolicy": "relay"},
|
151 |
media_stream_constraints={"video": True, "audio": False},
|
152 |
video_frame_callback=video_frame_callback,
|
153 |
async_processing=True,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
import time
|
2 |
|
3 |
# Logger Setup
|
|
|
10 |
|
11 |
# Initialize MediaPipe and Background Segmentor
|
12 |
mp_hands = mp.solutions.hands
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
["Z", "X", "C", "V", "B", "N", "M", ",", ".", "/"]]
|
14 |
|
15 |
class Button:
|
16 |
+
def __init__(self, pos, text, size=[100, 100]):
|
17 |
self.pos = pos
|
18 |
self.size = size
|
19 |
self.text = text
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
# Create the keyboard buttons
|
21 |
buttonList = []
|
22 |
h, w = img.shape[:2]
|
|
|
37 |
|
38 |
# Draw Keyboard Buttons
|
39 |
for button in buttonList:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
40 |
x4, y4 = int(hand_landmarks.landmark[mp_hands.HandLandmark.THUMB_TIP].x * w), int(hand_landmarks.landmark[mp_hands.HandLandmark.THUMB_TIP].y * h)
|
41 |
x8, y8 = int(hand_landmarks.landmark[mp_hands.HandLandmark.INDEX_FINGER_TIP].x * w), int(hand_landmarks.landmark[mp_hands.HandLandmark.INDEX_FINGER_TIP].y * h)
|
42 |
|
43 |
+
distance = np.sqrt((x8 - x4) ** 2 + (y8 - y4) ** 2)
|
44 |
+
click_threshold = 0.2*np.sqrt(bbox[2] ** 2 + bbox[3] ** 2)
|
45 |
|
46 |
for button in buttonList:
|
47 |
x, y = button.pos
|
|
|
48 |
if x < x8 < x + bw and y < y8 < y + bh:
|
49 |
cv2.rectangle(img, (x, y), (x + bw, y + bh), (0, 255, 160), -1)
|
50 |
cv2.putText(img, button.text, (x + int(0.2 * bw), y + int(0.7 * bh)), cv2.FONT_HERSHEY_PLAIN, font_scale, (255, 255, 255), font_thickness)
|
|
|
59 |
else:
|
60 |
output_text += ' ' # Add space
|
61 |
# Position and dimensions for the rectangle
|
62 |
+
|
63 |
text_x = int(0.05 * w)
|
64 |
text_y = int(0.70 * h)
|
65 |
text_width = int(0.9 * w) # Adjust width as needed
|
66 |
text_height = int(0.1 * h) # Adjust height as needed
|
67 |
# Draw the rectangle
|
68 |
+
|
69 |
cv2.rectangle(img,
|
70 |
(text_x, text_y - text_height), # Top-left corner
|
71 |
(text_x + text_width, text_y), # Bottom-right corner
|
|
|
81 |
return av.VideoFrame.from_ndarray(img, format="bgr24")
|
82 |
|
83 |
# WebRTC Streamer
|
|
|
|
|
|
|
|
|
84 |
media_stream_constraints={"video": True, "audio": False},
|
85 |
video_frame_callback=video_frame_callback,
|
86 |
async_processing=True,
|