File size: 1,975 Bytes
dc4f473
b1f6a8b
f31bdd1
 
 
b1f6a8b
f31bdd1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
67ddf1e
f31bdd1
 
 
 
 
 
 
 
 
 
 
 
 
 
67ddf1e
f31bdd1
 
93dd3e3
f31bdd1
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
import streamlit as st
from PIL import Image
from transformers import AutoFeatureExtractor, AutoModelForImageClassification
import torch
import requests
from io import BytesIO

# Load deepfake detection model and extractor
model_name = "systemkc/deepfake-detection-v1"  # Example model (replace if needed)
extractor = AutoFeatureExtractor.from_pretrained(model_name)
model = AutoModelForImageClassification.from_pretrained(model_name)

st.title("Deepfake Thumbnail Detector")
st.write("Upload a YouTube video link, and we’ll analyze the thumbnail to check for deepfakes.")

video_url = st.text_input("Enter YouTube Video Link:")
submit = st.button("Detect Deepfake")

if submit and video_url:
    try:
        video_id = video_url.split("v=")[-1].split("&")[0] if "v=" in video_url else video_url.split("/")[-1]
        thumbnail_url = f"https://img.youtube.com/vi/{video_id}/maxresdefault.jpg"
        response = requests.get(thumbnail_url)
        response.raise_for_status()

        image = Image.open(BytesIO(response.content))
        st.image(image, caption="Video Thumbnail", use_container_width=True)

        # Preprocess and predict
        inputs = extractor(images=image, return_tensors="pt")
        with torch.no_grad():
            outputs = model(**inputs)
            logits = outputs.logits
            predicted_class = torch.argmax(logits, dim=1).item()
            confidence = torch.softmax(logits, dim=1)[0, predicted_class].item()

        # Interpret results
        if predicted_class == 1:  # Assuming class 1 = Deepfake
            st.error(f"🚨 **Deepfake Detected** (Confidence: {confidence:.2%})")
            st.write("⚠️ Indicators of manipulation detected in the thumbnail.")
        else:
            st.success(f"✅ **No Deepfake Detected** (Confidence: {confidence:.2%})")
            st.write("👍 Thumbnail appears authentic based on the model’s analysis.")

    except Exception as e:
        st.error(f"Error: {str(e)}")