Update app.py

app.py

@@ -1,55 +1,39 @@
-import gradio as gr
+import cv2
 from deepface import DeepFace
-import tempfile
+import gradio as gr
 import os
 
-def analyze_video(video):
-    if video is None:
-        return "❌ No video file uploaded."
-
-    try:
-        # Save video to a temporary path
-        with tempfile.NamedTemporaryFile(delete=False, suffix=".mp4") as tmp:
-            tmp.write(video.read())
-            video_path = tmp.name
-
-        # Analyze video for emotion (can add other actions like age, gender, race)
-        result = DeepFace.analyze(
-            video_path,
-            actions=["emotion"],
-            detector_backend="retinaface",
-            enforce_detection=False
-        )
+def analyze_emotion_from_video(video_path):
+    if not video_path or not os.path.exists(video_path):
+        return "❌ Error: No video file found or path is invalid."
 
-        # Summarize emotion frequency
-        emotion_counts = {}
-        for frame in result:
-            emotion = frame.get("dominant_emotion")
-            if emotion:
-                emotion_counts[emotion] = emotion_counts.get(emotion, 0) + 1
+    cap = cv2.VideoCapture(video_path)
+    if not cap.isOpened():
+        return "❌ Error: Could not open the video file."
 
-        total_frames = sum(emotion_counts.values())
-        output = "📊 **Emotion Analysis:**\n"
-        for emotion, count in sorted(emotion_counts.items(), key=lambda x: x[1], reverse=True):
-            output += f"- {emotion}: {count} frames ({(count/total_frames)*100:.1f}%)\n"
+    frame_count = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
+    middle_frame_index = frame_count // 2
+    cap.set(cv2.CAP_PROP_POS_FRAMES, middle_frame_index)
+    success, frame = cap.read()
 
-        # Basic lie detection logic (can be customized)
-        if emotion_counts.get("fear", 0)/total_frames > 0.2 or emotion_counts.get("surprise", 0)/total_frames > 0.2:
-            output += "\n🔴 **Lie Detected: High fear/surprise levels**"
-        else:
-            output += "\n🟢 **No lie patterns detected**"
-
-        return output
+    if not success or frame is None:
+        return "❌ Error: Could not read a valid frame from the video."
 
+    try:
+        analysis = DeepFace.analyze(frame, actions=['emotion'], enforce_detection=False)
+        dominant_emotion = analysis[0]['dominant_emotion']
+        return f"✅ Dominant Emotion: {dominant_emotion}"
     except Exception as e:
         return f"❌ Error during analysis: {str(e)}"
-
-with gr.Blocks() as demo:
-    gr.Markdown("## 🕵️ Lie Detection from Video\nUpload a short face-visible video to detect emotional indicators.")
-    video_input = gr.Video(label="🎥 Upload MP4 Video")
-    output_box = gr.Textbox(label="📈 Results", lines=10)
-    analyze_btn = gr.Button("🔍 Analyze Video")
-
-    analyze_btn.click(fn=analyze_video, inputs=video_input, outputs=output_box)
-
-demo.launch()
+    finally:
+        cap.release()
+
+interface = gr.Interface(
+    fn=analyze_emotion_from_video,
+    inputs=gr.Video(label="Upload a Video File"),
+    outputs=gr.Textbox(label="Emotion Analysis Result"),
+    title="🎭 Video Emotion Detection",
+    description="Upload a short video with a visible face to detect the dominant emotion using DeepFace."
+)
+
+interface.launch()