added extract keyframes

app.py

@@ -1,32 +1,42 @@
 import gradio as gr
 from textblob import TextBlob
+import utils
+import tempfile
+
+
+def answer_video_question(query : str, url : str, file : bytes) -> dict:
+    # Either `file` or `url` must be provided
+    if file is not None:
+        with tempfile.NamedTemporaryFile(delete=False, suffix=".mp4") as temp_vid:
+            temp_vid.write(file.read())
+            temp_video_path = temp_vid.name
+
+        # Output frame folder
+        frame_output_path = './cache/uploaded_video/frames'
+        extract_keyframes(temp_video_path)
+
+        return {"out_vid_frame_from_file":frame_output_path}
+
+    elif url:
+        files_path = download_video(url)
+        extract_keyframes(files_path['video_path'])
+        return {"out_vid_path_from_url":}
+
+    else:
+        return {"error": "Please provide a movie file or URL."}
 
-def sentiment_analysis(text: str) -> dict:
-    """
-    Analyze the sentiment of the given text.
-
-    Args:
-        text (str): The text to analyze
-
-    Returns:
-        dict: A dictionary containing polarity, subjectivity, and assessment
-    """
-    blob = TextBlob(text)
-    sentiment = blob.sentiment
-    
-    return {
-        "polarity": round(sentiment.polarity, 2),  # -1 (negative) to 1 (positive)
-        "subjectivity": round(sentiment.subjectivity, 2),  # 0 (objective) to 1 (subjective)
-        "assessment": "positive" if sentiment.polarity > 0 else "negative" if sentiment.polarity < 0 else "neutral"
-    }
 
 # Create the Gradio interface
 demo = gr.Interface(
-    fn=sentiment_analysis,
-    inputs=gr.Textbox(placeholder="Enter text to analyze..."),
+    fn=answer_video_question,
+    inputs=[
+        gr.Textbox(placeholder="Enter Query about the movie", label="Query"),
+        gr.Textbox(placeholder="Paste the URL of the movie", label="Movie URL (optional)"),
+        gr.File(label="Upload Movie File (optional)")
+    ],
     outputs=gr.JSON(),
     title="Text Sentiment Analysis",
-    description="Analyze the sentiment of text using TextBlob"
+    description="Ask a question about a movie. You can provide a movie via a URL or by uploading a file. The movie will be cached and deleted when the Space goes to sleep."
 )
 
 # Launch the interface and MCP server

utils.py

@@ -63,4 +63,65 @@ def download_video(url):
     return {
         "video_path": video_filename,
         "audio_path": audio_filename,
-    }
+    }
+
+
+
+import cv2
+import os
+from skimage.metrics import structural_similarity as ssim
+
+
+def is_significantly_different(img1, img2, threshold=0.4):
+    """Check if two images are significantly different using SSIM.
+    Args:
+        img1 (numpy.ndarray): First image.
+        img2 (numpy.ndarray): Second image.
+        threshold (float): SSIM threshold to determine significant difference.
+    Returns:
+        bool: True if images are significantly different, False otherwise.
+    """
+    grayA = cv2.cvtColor(img1, cv2.COLOR_BGR2GRAY)
+    grayB = cv2.cvtColor(img2, cv2.COLOR_BGR2GRAY)
+    score, _ = ssim(grayA, grayB, full=True)
+    return score < threshold  # Lower score means more different
+
+
+
+def extract_keyframes(video_path, diff_threshold=0.4):
+    """Extract key frames from a video based on significant differences.
+    Args:
+        video_path (str): Path to the input video file.
+        output_path (str): Directory to save the extracted key frames.
+        diff_threshold (float): SSIM threshold to determine significant difference.
+    """
+    cap = cv2.VideoCapture(video_path)
+    frame_id = 0
+    saved_id = 0
+    success, prev_frame = cap.read()
+
+    if not success:
+        print("Failed to read video.")
+        return
+    
+    output_path='./cache/video/frames'
+    if not os.path.exists(output_path):
+        os.mkdir(output_path)
+
+    while True:
+        success, frame = cap.read()
+        if not success:
+            break
+        frame_id += 1
+
+        if is_significantly_different(prev_frame, frame, threshold=diff_threshold):
+            filename = os.path.join("./cache/video/frames/",f"keyframe_{saved_id:04d}.jpg")
+            cv2.imwrite(filename, frame)
+            prev_frame = frame
+            saved_id += 1
+
+    cap.release()
+    print(f"Extracted {saved_id} key frames.")
+
+# Example usage
+extract_keyframes(video_path)