Update app.py

app.py

@@ -3,10 +3,16 @@
 # Webcam → geometric detector → static WAV alert (with cooldown)
 # Live console logs of per-frame latency + status.
 # ──────────────────────────────────────────────────────────
-import time, os, yaml, logging, numpy as np, gradio as gr, soundfile as sf
-from dotenv import load_dotenv
+import time
+import os
+import yaml
+import logging
+import numpy as np
+import gradio as gr
+import soundfile as sf
 
-from src.detection.factory import get_detector    # your existing factory
+from dotenv import load_dotenv
+from src.detection.factory import get_detector  # your existing factory
 
 # ───────────────────────────── logging
 logging.basicConfig(
@@ -23,31 +29,36 @@ with open("config.yaml") as f:
 detector = get_detector(CFG)
 
 # ───────────────────────────── alert sound (read once)
-wav_path = CFG["alerting"]["alert_sound_path"]
+wav_path = CFG["alerting"].get("alert_sound_path", "alert.wav")
 try:
-    ALERT_SR, ALERT_DATA = sf.read(wav_path, dtype="float32")  #  (sr, np.ndarray)
+    ALERT_SR, ALERT_DATA = sf.read(wav_path, dtype="float32")
     logging.info(f"Loaded alert sound: {wav_path}  ({len(ALERT_DATA)/ALERT_SR:.2f}s)")
 except Exception as e:
     ALERT_SR, ALERT_DATA = None, None
     logging.warning(f"Failed to load alert sound: {e}")
 
-# ───────────────────────────── simple cooldown
-ALERT_COOLDOWN = CFG["alerting"].get("alert_cooldown_seconds", 7)
-_last_alert_ts = 0.0
-
-# ───────────────────────────── frame callback
-def process_live_frame(frame):
-    global _last_alert_ts
-
+# ───────────────────────────── frame processing
+def process_live_frame(frame, last_alert_ts):
     if frame is None:
-        return np.zeros((480, 640, 3), np.uint8), "Status: Inactive", None
+        return (
+            np.zeros((480, 640, 3), dtype=np.uint8),
+            "Status: Inactive",
+            None,
+            last_alert_ts
+        )
 
     t0 = time.perf_counter()
 
-    processed, indic, _ = detector.process_frame(frame)
-    level    = indic.get("drowsiness_level", "Awake")
+    try:
+        processed, indic, _ = detector.process_frame(frame)
+    except Exception as e:
+        logging.error(f"Error processing frame: {e}")
+        processed = np.zeros_like(frame)
+        indic = {"drowsiness_level": "Error", "lighting": "Unknown", "details": {"Score": 0.0}}
+
+    level = indic.get("drowsiness_level", "Awake")
     lighting = indic.get("lighting", "Good")
-    score    = indic.get("details", {}).get("Score", 0.0)
+    score = indic.get("details", {}).get("Score", 0.0)
 
     dt_ms = (time.perf_counter() - t0) * 1000.0
     logging.info(f"{dt_ms:6.1f} ms │ {lighting:<4} │ {level:<14} │ score={score:.2f}")
@@ -58,18 +69,20 @@ def process_live_frame(frame):
            else f"Status: {level}\nScore: {score:.2f}")
     )
 
-    # decide whether to play the alert
     audio_out = None
+    new_last_alert_ts = last_alert_ts
+    ALERT_COOLDOWN = CFG["alerting"].get("alert_cooldown_seconds", 7)
+
     if (
         ALERT_DATA is not None
         and level != "Awake"
         and lighting != "Low"
-        and (time.time() - _last_alert_ts) > ALERT_COOLDOWN
+        and (time.monotonic() - last_alert_ts) > ALERT_COOLDOWN
     ):
-        _last_alert_ts = time.time()
-        audio_out = (ALERT_SR, ALERT_DATA.copy())     # hand a fresh copy to Gradio
+        new_last_alert_ts = time.monotonic()
+        audio_out = (ALERT_SR, ALERT_DATA.copy())
 
-    return processed, status_txt, audio_out
+    return processed, status_txt, audio_out, new_last_alert_ts
 
 # ───────────────────────────── UI
 with gr.Blocks(title="Drive Paddy – Drowsiness Detection") as app:
@@ -80,19 +93,22 @@ with gr.Blocks(title="Drive Paddy – Drowsiness Detection") as app:
         with gr.Column(scale=2):
             cam = gr.Image(sources=["webcam"], streaming=True, label="Live Camera Feed")
         with gr.Column(scale=1):
-            out_img  = gr.Image(label="Processed Feed")
+            out_img = gr.Image(label="Processed Feed")
             out_text = gr.Textbox(label="Live Status", lines=3, interactive=False)
             out_audio = gr.Audio(
                 label="Alert",
                 autoplay=True,
-                type="numpy",        # expects (sr, np.ndarray)
+                type="numpy",
                 visible=True,
             )
+            last_alert_state = gr.State(value=0.0)
 
-    cam.stream(fn=process_live_frame,
-               inputs=cam,
-               outputs=[out_img, out_text, out_audio])
+    cam.stream(
+        fn=process_live_frame,
+        inputs=[cam, last_alert_state],
+        outputs=[out_img, out_text, out_audio, last_alert_state]
+    )
 
 if __name__ == "__main__":
     logging.info("Launching Gradio app …")
-    app.launch(debug=True)
+    app.launch(debug=True)