Update app.py

app.py

@@ -1,42 +1,93 @@
-# app.py
 import gradio as gr
-from transformers import pipeline
+import torch
+import cv2
+from transformers import AutoModelForImageClassification, AutoImageProcessor
+from PIL import Image
+import numpy as np
 
-# Load the deepfake detection model from Hugging Face
-detector = pipeline(
-    "video-classification",
-    model="Hemgg/deepfake-vs-real-video-detection"
+# ----------------------------------------------------------
+# Load Hugging Face GenConViT Model
+# ----------------------------------------------------------
+model = AutoModelForImageClassification.from_pretrained(
+    "Thanuja2109/GenConViT"
 )
 
-def analyze_video(video_path):
-    """
-    Run deepfake detection on uploaded video.
-    Returns:
-        label: predicted class ("FAKE" or "REAL")
-        score: confidence of the prediction
-        all_scores: full output from the model
-    """
-    results = detector(video_path)
-    top = results[0]  # assume the first element is top prediction
-    return top["label"], float(top["score"]), results
-
-# Gradio UI
-with gr.Blocks() as demo:
-    gr.Markdown("# DeepFake Video Detector")
-    gr.Markdown("Upload a video and the model will classify it as REAL or FAKE.")
-
-    video_input = gr.Video(label="Upload video")
-    analyze_btn = gr.Button("Analyze Video")
-
-    output_label = gr.Textbox(label="Predicted Label")
-    output_score = gr.Number(label="Confidence")
-    output_all = gr.JSON(label="All Predictions")
-
-    analyze_btn.click(
-        fn=analyze_video,
-        inputs=[video_input],
-        outputs=[output_label, output_score, output_all]
-    )
-
-if __name__ == "__main__":
-    demo.launch(server_name="0.0.0.0", share=True)
+processor = AutoImageProcessor.from_pretrained(
+    "Thanuja2109/GenConViT"
+)
+
+model.eval()
+
+# ----------------------------------------------------------
+# Deepfake detection function
+# ----------------------------------------------------------
+def detect_deepfake(video):
+
+    # Load video
+    cap = cv2.VideoCapture(video)
+
+    if not cap.isOpened():
+        return "Error: cannot open video", None
+
+    scores = []
+    frames_collected = 0
+
+    # Sample 1 frame every 10
+    frame_interval = 10
+
+    frame_img = None
+
+    i = 0
+    while True:
+        ret, frame = cap.read()
+        if not ret:
+            break
+
+        if i % frame_interval == 0:
+            # Convert to RGB
+            rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
+            pil_img = Image.fromarray(rgb)
+
+            inputs = processor(images=pil_img, return_tensors="pt")
+
+            with torch.no_grad():
+                logits = model(**inputs).logits
+                prob_fake = torch.softmax(logits, dim=1)[0][1].item()
+
+            scores.append(prob_fake)
+            frame_img = pil_img  # save last sampled frame
+
+        i += 1
+
+    cap.release()
+
+    if len(scores) == 0:
+        return "No frames processed", None
+
+    avg_score = np.mean(scores)
+
+    label = "🔴 Deepfake" if avg_score > 0.5 else "🟢 Real"
+
+    result_text = f"""
+### Prediction: **{label}**
+**Confidence (fake probability): {avg_score:.4f}**
+"""
+
+    return result_text, frame_img
+
+
+# ----------------------------------------------------------
+# Gradio Interface
+# ----------------------------------------------------------
+app = gr.Interface(
+    fn=detect_deepfake,
+    inputs=gr.Video(label="Upload a video"),
+    outputs=[
+        gr.Markdown(label="Prediction"),
+        gr.Image(label="Analyzed Frame")
+    ],
+    title="GenConViT Deepfake Video Detector",
+    description="Upload a video. The app samples frames and uses GenConViT to detect deepfakes."
+)
+
+app.launch()