Update app.py

app.py

@@ -1,106 +1,98 @@
 # app_gradio.py
-import gradio as gr
-import numpy as np
-import torch
-import os, yaml, soundfile as sf
-from dotenv import load_dotenv
-from threading import Thread
-import logging
-import time
-
-
-# --- TTS & AI Imports ---
-# from parler_tts import ParlerTTSForConditionalGeneration
-# from transformers import AutoTokenizer, AutoFeatureExtractor
-# from streamer import ParlerTTSStreamer                     # local file
-
-from src.detection.factory import get_detector
-
 # ──────────────────────────────────────────────────────────
-# CONFIG & BACKEND SET-UP
+# Webcam → geometric detector → static WAV alert (with cooldown)
+# Live console logs of per-frame latency + status.
 # ──────────────────────────────────────────────────────────
-load_dotenv()
+import time, os, yaml, logging, numpy as np, gradio as gr, soundfile as sf
+from dotenv import load_dotenv
+
+from src.detection.factory import get_detector    # your existing factory
 
+# ───────────────────────────── logging
 logging.basicConfig(
     level=logging.INFO,
-    format="%(asctime)s %(levelname)s │ %(message)s",
+    format="%(asctime)s │ %(message)s",
     datefmt="%H:%M:%S",
 )
 
+# ───────────────────────────── config / detector
+load_dotenv()
+with open("config.yaml") as f:
+    CFG = yaml.safe_load(f)
 
-with open("config.yaml", "r") as f:
-    config = yaml.safe_load(f)
-
-
-
-print("Initializing detector and alerter …")
-detector = get_detector(config)
-
-path = config["alerting"]["alert_sound_path"]
-with open(path, "rb") as f:
-    audio_bytes = f.read()
-print(f"[AlertSystem] loaded sound: {audio_bytes}")
-
-
-if audio_bytes is None:
-    logging.warning("No alert sound loaded; driver will not hear any audio!")
+detector = get_detector(CFG)
 
-print("Backend ready.")
+# ───────────────────────────── alert sound (read once)
+wav_path = CFG["alerting"]["alert_sound_path"]
+try:
+    ALERT_SR, ALERT_DATA = sf.read(wav_path, dtype="float32")  #  (sr, np.ndarray)
+    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 PROCESSOR
-# ──────────────────────────────────────────────────────────
+# ───────────────────────────── frame callback
 def process_live_frame(frame):
+    global _last_alert_ts
+
     if frame is None:
         return np.zeros((480, 640, 3), np.uint8), "Status: Inactive", None
 
-    t0 = time.time()
-    
-    processed, indicators, _ = detector.process_frame(frame)
-    level      = indicators.get("drowsiness_level", "Awake")
-    lighting   = indicators.get("lighting", "Good")
-    score      = indicators.get("details", {}).get("Score", 0)
+    t0 = time.perf_counter()
 
-    dt_ms = (time.time() - t0) * 1000.0
+    processed, indic, _ = detector.process_frame(frame)
+    level    = indic.get("drowsiness_level", "Awake")
+    lighting = indic.get("lighting", "Good")
+    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}")
 
-    status_txt = f"Lighting: {lighting}\n"
-    status_txt += ("Detection paused due to low light."
-                   if lighting == "Low"
-                   else f"Status: {level}\nScore: {score:.2f}")
+    status_txt = (
+        f"Lighting: {lighting}\n"
+        + ("Detection paused – low light." if lighting == "Low"
+           else f"Status: {level}\nScore: {score:.2f}")
+    )
 
-    if level != "Awake" and lighting != "Low":
-        audio_bytes = audio_bytes
-        logging.info(f"Printing {audio_bytes}")
-        if audio_bytes:
-            audio_out = audio_bytes
+    # decide whether to play the alert
+    audio_out = None
+    if (
+        ALERT_DATA is not None
+        and level != "Awake"
+        and lighting != "Low"
+        and (time.time() - _last_alert_ts) > ALERT_COOLDOWN
+    ):
+        _last_alert_ts = time.time()
+        audio_out = (ALERT_SR, ALERT_DATA.copy())     # hand a fresh copy to Gradio
 
     return processed, status_txt, audio_out
 
-# ──────────────────────────────────────────────────────────
-# GRADIO UI
-# ──────────────────────────────────────────────────────────
-with gr.Blocks(theme=gr.themes.Default(primary_hue="blue")) as app:
-    gr.Markdown("# 🚗 Drive Paddy – Drowsiness Detection")
-    gr.Markdown("Live detection with real-time voice alerts.")
+# ───────────────────────────── UI
+with gr.Blocks(title="Drive Paddy – Drowsiness Detection") as app:
+    gr.Markdown("# 🚗 **Drive Paddy** – Static-file Alert Demo")
+    gr.Markdown("Webcam-based drowsiness detection · console shows real-time logs.")
 
     with gr.Row():
         with gr.Column(scale=2):
-            webcam = gr.Image(sources=["webcam"], streaming=True,
-                              label="Live Camera Feed")
+            cam = gr.Image(sources=["webcam"], streaming=True, label="Live Camera Feed")
         with gr.Column(scale=1):
-            processed_img = gr.Image(label="Processed Feed")
-            status_box    = gr.Textbox(label="Live Status", lines=3, interactive=False)
-            alert_audio   = gr.Audio(label="Alert",
-                                     autoplay=True,
-                                     streaming=True)
-
-    webcam.stream(
-        fn=process_live_frame,
-        inputs=webcam,
-        outputs=[processed_img, status_box, alert_audio],
-    )
+            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)
+                visible=True,
+                height=60,
+            )
+
+    cam.stream(fn=process_live_frame,
+               inputs=cam,
+               outputs=[out_img, out_text, out_audio])
 
 if __name__ == "__main__":
     logging.info("Launching Gradio app …")