Update app.py

app.py

@@ -1,3 +1,6 @@
+import os
+os.environ["MEDIAPIPE_DISABLE_GPU"] = "1"
+
 import gradio as gr
 import cv2
 import numpy as np
@@ -5,11 +8,7 @@ import json
 import time
 from gtts import gTTS
 import tempfile
-import requests
 from PIL import Image
-from io import BytesIO
-
-# ✅ FIXED MEDIAPIPE IMPORT (HF SAFE)
 import mediapipe as mp
 from mediapipe import solutions
 
@@ -17,41 +16,39 @@ from mediapipe import solutions
 with open("gestures_rules.json", "r") as f:
     gesture_data = json.load(f)["gestures"]
 
-# ---------------- MEDIAPIPE SETUP ----------------
+# ---------------- MEDIAPIPE (CPU ONLY) ----------------
 mp_hands = solutions.hands
 hands = mp_hands.Hands(
+    static_image_mode=False,
     max_num_hands=1,
     min_detection_confidence=0.7,
     min_tracking_confidence=0.7
 )
 
-# ---------------- UTIL FUNCTIONS ----------------
+# ---------------- UTIL ----------------
 def get_finger_states(hand_landmarks):
     tips = [4, 8, 12, 16, 20]
     pips = [2, 6, 10, 14, 18]
-
-    states = []
-    for tip, pip in zip(tips, pips):
-        states.append(
-            1 if hand_landmarks.landmark[tip].y <
-            hand_landmarks.landmark[pip].y else 0
-        )
-    return states
+    return [
+        1 if hand_landmarks.landmark[t].y <
+        hand_landmarks.landmark[p].y else 0
+        for t, p in zip(tips, pips)
+    ]
 
 def detect_gesture(states):
-    for name, rule in gesture_data.items():
-        if rule["pattern"] == states:
-            return name
+    for k, v in gesture_data.items():
+        if v["pattern"] == states:
+            return k
     return None
 
 def speak_text(text):
     tts = gTTS(text=text)
-    temp = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3")
-    tts.save(temp.name)
-    return temp.name
+    f = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3")
+    tts.save(f.name)
+    return f.name
 
-# ---------------- FRAME PROCESSING ----------------
-def process_frame(frame, sentence, last_char, last_time):
+# ---------------- VIDEO PROCESS ----------------
+def process_video(frame, sentence, last_char, last_time):
     if frame is None:
         return frame, sentence, last_char, last_time
 
@@ -65,108 +62,57 @@ def process_frame(frame, sentence, last_char, last_time):
         char = detect_gesture(states)
 
         now = time.time()
-        if (
-            char
-            and char != last_char
-            and now - last_time > 1
-            and len(char) == 1
-        ):
+        if char and char != last_char and now - last_time > 1:
             sentence += char
             last_char = char
             last_time = now
 
     return frame, sentence, last_char, last_time
 
-# ---------------- BUTTON ACTIONS ----------------
 def clear_text():
     return "", "", 0.0
 
 def speak(sentence):
-    if sentence.strip():
-        return speak_text(sentence)
-    return None
+    return speak_text(sentence) if sentence else None
 
-# ---------------- LOAD LOGO ----------------
-logo_url = "https://raw.githubusercontent.com/imarshbir/Hand2Voice/main/logo.png"
-logo = Image.open(BytesIO(requests.get(logo_url).content))
-
-# ---------------- LOAD CSS ----------------
+# ---------------- CSS ----------------
 with open("styles.css") as f:
     custom_css = f.read()
 
 # ---------------- UI ----------------
-with gr.Blocks(css=custom_css, title="Hand2Voice") as demo:
-
-    gr.Image(
-        value=logo,
-        show_label=False,
-        height=100
-    )
+with gr.Blocks(title="Hand2Voice") as demo:
 
-    gr.HTML("""
-    <div class='mission-box'>
-        <div class='mission-title'>Bridging Communication Gaps with AI</div>
-        <div class='mission-text'>
-            An intelligent system converting hand gestures into speech
-            in real time for inclusive communication.
-        </div>
-    </div>
-    """)
+    gr.Markdown("## 🤟 Hand2Voice – Gesture to Speech")
 
     with gr.Row():
-        with gr.Column(scale=1.1):
-            gr.Markdown("### 🎥 Live Camera Feed")
-            webcam = gr.Image(
-                source="webcam",
+        with gr.Column(scale=1):
+            webcam = gr.Video(
+                sources=["webcam"],
                 streaming=True,
-                type="numpy"
+                height=360
             )
 
         with gr.Column(scale=1):
-            gr.Markdown("### 📝 Translation Output")
-            output_box = gr.HTML(
-                "<div class='output-text'>🤚 Waiting for gestures...</div>"
-            )
+            output = gr.HTML("<h3>Waiting for gestures...</h3>")
+            speak_btn = gr.Button("🔊 Speak")
+            clear_btn = gr.Button("🧹 Clear")
+            audio = gr.Audio(autoplay=True)
 
-            speak_btn = gr.Button("🔊 Speak Sentence")
-            clear_btn = gr.Button("🧹 Clear Text")
-            audio_out = gr.Audio(autoplay=True)
+    sentence = gr.State("")
+    last_char = gr.State("")
+    last_time = gr.State(0.0)
 
-    # ---------------- STATES ----------------
-    sentence_state = gr.State("")
-    last_char_state = gr.State("")
-    last_time_state = gr.State(0.0)
-
-    # ---------------- STREAM ----------------
     webcam.stream(
-        process_frame,
-        inputs=[webcam, sentence_state, last_char_state, last_time_state],
-        outputs=[webcam, sentence_state, last_char_state, last_time_state]
+        process_video,
+        inputs=[webcam, sentence, last_char, last_time],
+        outputs=[webcam, sentence, last_char, last_time]
     ).then(
-        lambda s: f"<div class='output-text'>{s}</div>",
-        inputs=sentence_state,
-        outputs=output_box
-    )
-
-    # ---------------- BUTTONS ----------------
-    clear_btn.click(
-        clear_text,
-        outputs=[sentence_state, last_char_state, last_time_state]
-    )
-
-    speak_btn.click(
-        speak,
-        inputs=sentence_state,
-        outputs=audio_out
+        lambda s: f"<h2>{s}</h2>",
+        sentence,
+        output
     )
 
-    # ---------------- FOOTER ----------------
-    gr.HTML("""
-    <div class='footer-container'>
-        <div class='footer-title'>Hand2Voice</div>
-        <div class='footer-tagline'>Powered by Computer Vision & NLP</div>
-        <div class='footer-tagline'>Made by Arshbir Singh</div>
-    </div>
-    """)
+    clear_btn.click(clear_text, outputs=[sentence, last_char, last_time])
+    speak_btn.click(speak, sentence, audio)
 
-demo.launch()
+demo.launch(css=custom_css)