Views
No views yet
| Metric | Score |
|---|---|
| Accuracy | 91.2% |
| F1 Score | 91.0% |
| Precision | 91.5% |
| Recall | 91.2% |
| AUC-ROC | 96.2% |
1from app import load_model, predict
2from PIL import Image
3
4# Load model
5load_model()
6
7# Predict
8image = Image.open("food.jpg")
9result = predict("Amazing pizza with perfect crust!", image)
10print(result)Input: (Review Text, Product Image)
│ │
▼ ▼
┌─────────┐ ┌──────────┐
│ BERT │ │ CLIP │
│ Encoder │ │ Vision │
└────┬────┘ └────┬─────┘
│ │
┌────┴────┐ ┌────┴─────┐
│ 768-dim │ │ 768-dim │
└────┬────┘ └────┬─────┘
│ │
┌────┴────┐ ┌────┴─────┐
│Text Proj│ │Image Proj│
│ Head │ │ Head │
└────┬────┘ └────┬─────┘
│ │
▼ ▼
┌─────────────────────┐
│ Consistency Score │ ← Cosine Similarity
│ (Contrastive Loss) │
└─────────────────────┘
│
▼
┌─────────────────────┐
│ Concatenation │
│ (1536-dim) │
└──────────┬──────────┘
│
┌──────────▼──────────┐
│ MLP Classifier │
│ 512 → 256 → 2 │
└──────────┬──────────┘
│
▼
Output: [Authentic, Fake]1@article{multimodal_fake_review_2026,
2 title={Multimodal Fake Review Detection Using Contrastive Learning:
3 Leveraging Text-Image Alignment for AI-Generated Content Identification},
4 author={Your Name},
5 year={2026}
6}