Spaces:
Sleeping
Sleeping
File size: 11,037 Bytes
e34b3e2 632a239 e34b3e2 632a239 e34b3e2 632a239 e34b3e2 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 | import cv2
import mediapipe as mp
import streamlit as st
import numpy as np
import time
import pandas as pd
import altair as alt
from scipy.signal import butter, filtfilt, find_peaks
# -------------------------------
# Constants
# -------------------------------
LEFT_EYE = [33, 160, 158, 133, 153, 144]
RIGHT_EYE = [362, 385, 387, 263, 373, 380]
FOREHEAD_ROI = [10, 109, 67, 103, 54, 21, 162, 127, 234, 93, 132, 58, 172, 136, 150, 149, 176, 148]
BUFFER_SIZE = 300 # ~10s if 30 FPS
EAR_THRESHOLD = 0.22
DROWSINESS_TIME_THRESHOLD = 2.0 # seconds
# -------------------------------
# Mediapipe face mesh
# -------------------------------
mp_face_mesh = mp.solutions.face_mesh
# -------------------------------
# Helper Functions
# -------------------------------
def get_eye_aspect_ratio(landmarks, eye_indices):
"""Calculates the Eye Aspect Ratio (EAR) for a single eye."""
pts = np.array([(landmarks[i].x, landmarks[i].y) for i in eye_indices])
A = np.linalg.norm(pts[1] - pts[5])
B = np.linalg.norm(pts[2] - pts[4])
C = np.linalg.norm(pts[0] - pts[3])
ear = (A + B) / (2.0 * C)
return ear
def bandpass_filter(data, low=0.8, high=2.5, fs=30):
"""Applies a bandpass filter to the signal."""
nyq = 0.5 * fs
b, a = butter(1, [low/nyq, high/nyq], btype="band")
return filtfilt(b, a, data)
def compute_hr_hrv(signal, times, fs=30):
"""Computes Heart Rate (HR) and Heart Rate Variability (HRV) from the rPPG signal."""
if len(signal) < fs * 3: # need at least 3s of data
return None, None
try:
# Detrending the signal to remove baseline wander
signal_detrended = signal - np.mean(signal)
filtered = bandpass_filter(signal_detrended, fs=fs)
# Using a more robust peak finding
peaks, properties = find_peaks(filtered, distance=fs*0.7, prominence=np.std(filtered)*0.3)
if len(peaks) < 3: # Need at least 3 peaks for a more stable HR
return None, None
peak_times = np.array(times)[peaks]
rr_intervals = np.diff(peak_times) # in seconds
# Basic outlier removal for RR intervals
median_rr = np.median(rr_intervals)
valid_rr = rr_intervals[np.abs(rr_intervals - median_rr) < 0.3 * median_rr]
if len(valid_rr) < 2:
return None, None
hr = 60.0 / np.mean(valid_rr)
hrv = np.std(valid_rr) * 1000 # RMSSD is a better HRV metric, but this is a start
# Plausible HR range
if not (40 < hr < 160):
return None, None
return hr, hrv
except (np.linalg.LinAlgError, ValueError):
return None, None
def initialize_session_state():
"""Initializes Streamlit session state variables."""
if "history" not in st.session_state:
st.session_state.history = {
"time": [], "blink_rate": [], "hr": [], "hrv": []
}
if "blink_count" not in st.session_state:
st.session_state.blink_count = 0
if "last_eye_state" not in st.session_state:
st.session_state.last_eye_state = "open"
if "start_time" not in st.session_state:
st.session_state.start_time = time.time()
if "signal_buffer" not in st.session_state:
st.session_state.signal_buffer = []
if "time_buffer" not in st.session_state:
st.session_state.time_buffer = []
if "drowsy_start_time" not in st.session_state:
st.session_state.drowsy_start_time = None
def update_dashboard(placeholders, data):
"""Updates the Streamlit dashboard with new data."""
placeholders["stframe"].image(data["frame"], channels="BGR")
# --- Metrics ---
placeholders["metrics"]["blink"].metric("Blink Rate (per min)", f"{data['blink_rate']:.2f}")
placeholders["metrics"]["hr"].metric("Heart Rate (bpm)", f"{data['hr']:.1f}" if data['hr'] is not None else "N/A")
placeholders["metrics"]["hrv"].metric("HRV (ms)", f"{data['hrv']:.1f}" if data['hrv'] is not None else "N/A")
# --- Drowsiness Alert ---
if data["drowsy_alert"]:
placeholders["alert"].warning("🚨 Drowsiness Detected!")
else:
placeholders["alert"].empty()
# --- Charts ---
df = pd.DataFrame(st.session_state.history)
df["time"] = pd.to_datetime(df["time"], unit="s")
with placeholders["charts"]["blink_tab"]:
chart = alt.Chart(df).mark_line().encode(
x=alt.X('time:T', title='Time'),
y=alt.Y('blink_rate:Q', title='Blink Rate (per min)')
).properties(title="Blink Rate Over Time")
placeholders["charts"]["blink_chart"].altair_chart(chart, use_container_width=True)
with placeholders["charts"]["hr_tab"]:
chart = alt.Chart(df).mark_line().encode(
x=alt.X('time:T', title='Time'),
y=alt.Y('hr:Q', title='Heart Rate (bpm)')
).properties(title="Heart Rate Over Time")
placeholders["charts"]["hr_chart"].altair_chart(chart, use_container_width=True)
with placeholders["charts"]["hrv_tab"]:
chart = alt.Chart(df).mark_line().encode(
x=alt.X('time:T', title='Time'),
y=alt.Y('hrv:Q', title='HRV (ms)')
).properties(title="HRV Over Time")
placeholders["charts"]["hrv_chart"].altair_chart(chart, use_container_width=True)
def main():
"""Main function to run the Streamlit application."""
st.set_page_config(page_title="DriFit - Driver Monitoring", layout="wide")
st.title("DriFit: In-Car Driver Health & Fatigue Monitoring")
st.info("This application uses your webcam to monitor driver fatigue and health metrics in real-time.")
initialize_session_state()
# --- UI Placeholders ---
col1, col2 = st.columns([2, 1])
with col1:
stframe = st.empty()
alert_placeholder = st.empty()
with col2:
m_col1, m_col2, m_col3 = st.columns(3)
st.subheader("Metrics")
blink_metric_placeholder = m_col1.empty()
hr_metric_placeholder = m_col2.empty()
hrv_metric_placeholder = m_col3.empty()
st.subheader("Metrics Over Time")
blink_tab, hr_tab, hrv_tab = st.tabs(["Blink Rate", "Heart Rate", "HRV"])
with blink_tab:
blink_chart_placeholder = st.empty()
with hr_tab:
hr_chart_placeholder = st.empty()
with hrv_tab:
hrv_chart_placeholder = st.empty()
placeholders = {
"stframe": stframe,
"alert": alert_placeholder,
"metrics": {"blink": blink_metric_placeholder, "hr": hr_metric_placeholder, "hrv": hrv_metric_placeholder},
"charts": {
"blink_tab": blink_tab, "hr_tab": hr_tab, "hrv_tab": hrv_tab,
"blink_chart": blink_chart_placeholder,
"hr_chart": hr_chart_placeholder,
"hrv_chart": hrv_chart_placeholder
}
}
# --- Webcam and Face Mesh ---
if 'cap' not in st.session_state:
st.session_state.cap = cv2.VideoCapture(0)
if 'face_mesh' not in st.session_state:
st.session_state.face_mesh = mp_face_mesh.FaceMesh(refine_landmarks=True)
cap = st.session_state.cap
face_mesh = st.session_state.face_mesh
run = st.checkbox('Run')
if not cap.isOpened():
st.error("Could not open webcam. Please grant access and refresh.")
return
while run:
ret, frame = cap.read()
if not ret:
st.warning("Could not read frame from webcam. Stopping.")
run = False
break
frame = cv2.flip(frame, 1)
rgb_frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
results = face_mesh.process(rgb_frame)
eye_state = "open"
hr, hrv = None, None
drowsy_alert = False
if results.multi_face_landmarks:
face_landmarks = results.multi_face_landmarks[0]
# --- Eye tracking (fatigue) ---
left_ear = get_eye_aspect_ratio(face_landmarks.landmark, LEFT_EYE)
right_ear = get_eye_aspect_ratio(face_landmarks.landmark, RIGHT_EYE)
avg_ear = (left_ear + right_ear) / 2.0
if avg_ear < EAR_THRESHOLD:
eye_state = "closed"
if st.session_state.drowsy_start_time is None:
st.session_state.drowsy_start_time = time.time()
elif time.time() - st.session_state.drowsy_start_time > DROWSINESS_TIME_THRESHOLD:
drowsy_alert = True
else:
eye_state = "open"
st.session_state.drowsy_start_time = None
if st.session_state.last_eye_state == "closed" and eye_state == "open":
st.session_state.blink_count += 1
st.session_state.last_eye_state = eye_state
# --- rPPG HR & HRV (forehead ROI) ---
h, w, _ = frame.shape
forehead_pts = np.array([(face_landmarks.landmark[i].x * w, face_landmarks.landmark[i].y * h) for i in FOREHEAD_ROI], dtype=np.int32)
mask = np.zeros(frame.shape[:2], dtype=np.uint8)
cv2.fillConvexPoly(mask, forehead_pts, 255)
roi = cv2.bitwise_and(frame, frame, mask=mask)
x, y, w_roi, h_roi = cv2.boundingRect(forehead_pts)
if w_roi > 0 and h_roi > 0:
roi_cropped = roi[y:y+h_roi, x:x+w_roi]
if roi_cropped.size > 0:
green_mean = np.mean(roi_cropped[:, :, 1])
st.session_state.signal_buffer.append(green_mean)
st.session_state.time_buffer.append(time.time())
if len(st.session_state.signal_buffer) > BUFFER_SIZE:
st.session_state.signal_buffer.pop(0)
st.session_state.time_buffer.pop(0)
hr, hrv = compute_hr_hrv(st.session_state.signal_buffer, st.session_state.time_buffer)
cv2.polylines(frame, [forehead_pts], isClosed=True, color=(0, 255, 0), thickness=1)
# --- Data Update ---
elapsed = time.time() - st.session_state.start_time
blink_rate = (st.session_state.blink_count / (elapsed / 60)) if elapsed > 5 else 0.0
# Update history
st.session_state.history["time"].append(time.time())
st.session_state.history["blink_rate"].append(blink_rate)
st.session_state.history["hr"].append(hr)
st.session_state.history["hrv"].append(hrv)
for key in st.session_state.history:
st.session_state.history[key] = st.session_state.history[key][-100:]
# --- Dashboard Update ---
update_data = {
"frame": frame,
"blink_rate": blink_rate,
"hr": hr,
"hrv": hrv,
"drowsy_alert": drowsy_alert
}
update_dashboard(placeholders, update_data)
else:
if 'cap' in st.session_state:
st.session_state.cap.release()
del st.session_state.cap
if __name__ == "__main__":
main()
|