📊 Native vs Recaptured Classifier A deep learning-powered binary image classifier built with TensorFlow (MobileNetV2) and deployed using Streamlit. It predicts whether an image is Native (original) or Recaptured (re-captured from another device/screen). 🔗 Live : https://native-vs-recaptured-classifier-ygnlsemkvowpfc4jcy6j59.streamlit.app/ 🚀 Features

Upload multiple images simultaneously. Predicts probabilities for: prob_native: Likelihood of an image being original/native. prob_recaptured: Likelihood of an image being recaptured.

Generates a downloadable CSV file with predictions for all uploaded images. Uses MobileNetV2’s preprocess_input for consistent preprocessing. Optimized for fast inference with a lightweight model.

🛠️ Installation & Setup

1. Clone the Repository git clone https://github.com/Rijul-Tandon/native-vs-recaptured-classifier.git cd native-vs-recaptured-classifier

2. Create a Virtual Environment (Optional but Recommended) python -m venv venv

source venv/bin/activate

venv\Scripts\activate

3. Install Dependencies pip install -r requirements.txt

4. Run the Streamlit App Locally python -m streamlit run app.py

📂 Project Structure . ├── app.py # Streamlit web app ├── requirements.txt # Dependencies ├── Mewat/ │ └── model_checkpoints/ │ └── best_model.weights.h5 # Trained model weights └── README.md # Project documentation

📊 Example Output After uploading images, a table like the following is displayed:

Filename prob_native prob_recaptured

img1.png 0.82 0.18

img2.jpg 0.15 0.85

Results can be downloaded as a CSV file. 🧠 Model Details

Base Model: MobileNetV2 (pre-trained on ImageNet, frozen during training) Input Size: 224×224×3 Preprocessing: tf.keras.applications.mobilenet_v2.preprocess_input Output: Sigmoid layer providing the probability of an image being recaptured

🌐 Deployment Hosted on Streamlit Cloud:👉 Live Demo 📜 License This project is licensed under the MIT License. Feel free to use and modify for your work.