Update app.py
Browse files
app.py
CHANGED
|
@@ -10,18 +10,14 @@ app = Flask(__name__)
|
|
| 10 |
upload_folder = os.path.join('static', 'uploads')
|
| 11 |
os.makedirs(upload_folder, exist_ok=True)
|
| 12 |
|
| 13 |
-
# Fake News Detection
|
| 14 |
-
|
| 15 |
-
"mrm8488": pipeline("text-classification", model="mrm8488/bert-tiny-finetuned-fake-news-detection"),
|
| 16 |
-
"google-electra": pipeline("text-classification", model="google/electra-base-discriminator"),
|
| 17 |
-
"bert-base": pipeline("text-classification", model="bert-base-uncased")
|
| 18 |
-
}
|
| 19 |
|
| 20 |
# Image Detection Model (CLIP-based)
|
| 21 |
clip_model = CLIPModel.from_pretrained("openai/clip-vit-base-patch32")
|
| 22 |
clip_processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32")
|
| 23 |
|
| 24 |
-
# HTML Template with
|
| 25 |
HTML_TEMPLATE = """
|
| 26 |
<!DOCTYPE html>
|
| 27 |
<html lang="en">
|
|
@@ -42,19 +38,14 @@ HTML_TEMPLATE = """
|
|
| 42 |
<h1>📰 Fake News Detection</h1>
|
| 43 |
<form method="POST" action="/detect">
|
| 44 |
<textarea name="text" placeholder="Enter news text..." required></textarea>
|
| 45 |
-
<label for="model">Select Fake News Model:</label>
|
| 46 |
-
<select name="model" required>
|
| 47 |
-
<option value="mrm8488">MRM8488 (BERT-Tiny)</option>
|
| 48 |
-
<option value="google-electra">Google Electra (Base Discriminator)</option>
|
| 49 |
-
<option value="bert-base">BERT-Base Uncased</option>
|
| 50 |
-
</select>
|
| 51 |
<button type="submit">Detect News Authenticity</button>
|
| 52 |
</form>
|
| 53 |
|
| 54 |
{% if news_prediction %}
|
| 55 |
<div class="result">
|
| 56 |
<h2>🧠 News Detection Result:</h2>
|
| 57 |
-
<p>{{ news_prediction }}</p>
|
|
|
|
| 58 |
</div>
|
| 59 |
{% endif %}
|
| 60 |
|
|
@@ -68,7 +59,7 @@ HTML_TEMPLATE = """
|
|
| 68 |
<div class="result">
|
| 69 |
<h2>📷 Image Detection Result:</h2>
|
| 70 |
<p>{{ image_prediction|safe }}</p>
|
| 71 |
-
<p><strong>Explanation:</strong> The model compares the uploaded image against
|
| 72 |
</div>
|
| 73 |
{% endif %}
|
| 74 |
</div>
|
|
@@ -83,16 +74,17 @@ def home():
|
|
| 83 |
@app.route("/detect", methods=["POST"])
|
| 84 |
def detect():
|
| 85 |
text = request.form.get("text")
|
| 86 |
-
|
|
|
|
| 87 |
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
result = news_models[model_key](text)[0]
|
| 92 |
-
label = "REAL" if result['label'].lower() in ["real", "label_1", "neutral"] else "FAKE"
|
| 93 |
confidence = result['score'] * 100
|
| 94 |
|
| 95 |
-
prediction_text =
|
|
|
|
|
|
|
|
|
|
| 96 |
return render_template_string(HTML_TEMPLATE, news_prediction=prediction_text)
|
| 97 |
|
| 98 |
@app.route("/detect_image", methods=["POST"])
|
|
@@ -103,7 +95,6 @@ def detect_image():
|
|
| 103 |
file = request.files["image"]
|
| 104 |
img = Image.open(file).convert("RGB")
|
| 105 |
|
| 106 |
-
# Compare with AI and Human prompts
|
| 107 |
prompts = ["AI-generated image", "Human-created image"]
|
| 108 |
inputs = clip_processor(text=prompts, images=img, return_tensors="pt", padding=True)
|
| 109 |
|
|
|
|
| 10 |
upload_folder = os.path.join('static', 'uploads')
|
| 11 |
os.makedirs(upload_folder, exist_ok=True)
|
| 12 |
|
| 13 |
+
# Better Fake News Detection Model (DeBERTa for improved performance)
|
| 14 |
+
news_model = pipeline("text-classification", model="microsoft/deberta-v3-base-mnli")
|
|
|
|
|
|
|
|
|
|
|
|
|
| 15 |
|
| 16 |
# Image Detection Model (CLIP-based)
|
| 17 |
clip_model = CLIPModel.from_pretrained("openai/clip-vit-base-patch32")
|
| 18 |
clip_processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32")
|
| 19 |
|
| 20 |
+
# HTML Template with enhanced explanations
|
| 21 |
HTML_TEMPLATE = """
|
| 22 |
<!DOCTYPE html>
|
| 23 |
<html lang="en">
|
|
|
|
| 38 |
<h1>📰 Fake News Detection</h1>
|
| 39 |
<form method="POST" action="/detect">
|
| 40 |
<textarea name="text" placeholder="Enter news text..." required></textarea>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 41 |
<button type="submit">Detect News Authenticity</button>
|
| 42 |
</form>
|
| 43 |
|
| 44 |
{% if news_prediction %}
|
| 45 |
<div class="result">
|
| 46 |
<h2>🧠 News Detection Result:</h2>
|
| 47 |
+
<p>{{ news_prediction|safe }}</p>
|
| 48 |
+
<p><strong>Explanation:</strong> The model analyzes the text and classifies it as REAL or FAKE based on linguistic patterns, fact alignment, and contextual cues. High confidence indicates stronger evidence for the classification.</p>
|
| 49 |
</div>
|
| 50 |
{% endif %}
|
| 51 |
|
|
|
|
| 59 |
<div class="result">
|
| 60 |
<h2>📷 Image Detection Result:</h2>
|
| 61 |
<p>{{ image_prediction|safe }}</p>
|
| 62 |
+
<p><strong>Explanation:</strong> The model compares the uploaded image against prompts like "AI-generated image" and "Human-created image." Higher similarity to the AI prompt suggests an AI-generated image, and vice versa.</p>
|
| 63 |
</div>
|
| 64 |
{% endif %}
|
| 65 |
</div>
|
|
|
|
| 74 |
@app.route("/detect", methods=["POST"])
|
| 75 |
def detect():
|
| 76 |
text = request.form.get("text")
|
| 77 |
+
if not text:
|
| 78 |
+
return render_template_string(HTML_TEMPLATE, news_prediction="Invalid input.")
|
| 79 |
|
| 80 |
+
result = news_model(text)[0]
|
| 81 |
+
label = "REAL" if result['label'].lower() in ["entailment", "neutral"] else "FAKE"
|
|
|
|
|
|
|
|
|
|
| 82 |
confidence = result['score'] * 100
|
| 83 |
|
| 84 |
+
prediction_text = (
|
| 85 |
+
f"News is <strong>{label}</strong> (Confidence: {confidence:.2f}%)<br>"
|
| 86 |
+
f"Reasoning: The model evaluated the likelihood that the provided text aligns with factual information versus being contradictory or fabricated."
|
| 87 |
+
)
|
| 88 |
return render_template_string(HTML_TEMPLATE, news_prediction=prediction_text)
|
| 89 |
|
| 90 |
@app.route("/detect_image", methods=["POST"])
|
|
|
|
| 95 |
file = request.files["image"]
|
| 96 |
img = Image.open(file).convert("RGB")
|
| 97 |
|
|
|
|
| 98 |
prompts = ["AI-generated image", "Human-created image"]
|
| 99 |
inputs = clip_processor(text=prompts, images=img, return_tensors="pt", padding=True)
|
| 100 |
|