created a README file

README.md

@@ -0,0 +1,66 @@
+---
+title: DeBERTa+ViT FakeNews 
+emoji: 🐠
+colorFrom: pink
+colorTo: purple
+sdk: gradio
+sdk_version: 5.35.0
+app_file: app.py
+pinned: false
+license: mit
+---
+
+
+
+
+## Fake News Classifier (DeBERT+ViT)
+
+This project focuses on detecting fake news using a multimodal approach that combines a fine-tuned DeBERTa model for text analysis and a Vision Transformer (ViT) model for image classification.
+
+---
+
+## ⚠️Disclaimer
+
+- This project is for **educational and experimental purposes only**.
+- It is **not suitable for real-world fact-checking** or serious decision-making.
+- The system classifies content based on learned patterns, not actual truth or factual accuracy.
+
+---
+
+## Project Overview
+
+This fake news classifier was built as part of a research internship to:
+
+- This project explores fake news detection using a multimodal approach, combining both text and image analysis.
+- It utilizes DeBERTa for binary classification of news content and Vision Transformer (ViT) for evaluating associated images.
+- The system is trained on labeled datasets from Kaggle and Hugging Face, aiming to improve detection accuracy by analyzing both linguistic and visual patterns.
+
+---
+
+# 🧠 How it Works
+
+*DeBERTa Model*
+
+- A *DeBERTa-based model* was fine-tuned on labeled news text for fake news detection.
+- Input text was preprocessed and tokenized using DebertaTokenizer.
+- The model performs binary classification, where *0=Fake and 1=Real*
+
+*ViT Model*
+
+- A *Vision Transformer model* was fine-tuned on real vs fake news images using the vit-base-patch16-224-in21k checkpoint.
+- Images were preprocessed into patches using ViTImageProcessor and passed through the model for binary classification.
+- Outputs are binary class predictions where *0 = Fake, 1 = Real*.
+
+---
+
+## 🛠Libraries Used
+
+- `transformers`
+- `datasets`
+- `torch`
+- `scikit-learn`
+- `pandas`
+- `nltk` (optional preprocessing)
+- TorchVision
+
+---