Bollywood Celebrity Classifier

This project is a Bollywood Celebrity Classifier that uses a pre-trained VGG16 model to identify which Bollywood celebrity an uploaded image resembles.

Table of Contents

• Introduction
• Features
• Dataset
• Model
• Feature Extraction
• Setup
• Usage
• License

Introduction

The Bollywood Celebrity Classifier is built using a Convolutional Neural Network (CNN) based on the VGG16 architecture. It takes an image as input and predicts the Bollywood celebrity that the image resembles the most.

Features

• Image Upload: Upload an image to the web app.
• Feature Extraction: Extract features from the uploaded image using VGG16.
• Similarity Measurement: Measure the cosine similarity between the uploaded image and a set of precomputed celebrity images.
• Prediction: Display the most similar celebrity image and name.

Dataset

The dataset consists of images of various Bollywood celebrities. The images are used to extract features and build a similarity-based classifier.
Link : https://www.kaggle.com/datasets/sushilyadav1998/bollywood-celeb-localized-face-dataset

Model

The project uses the VGG16 model pre-trained on the ImageNet dataset. The top layers of the VGG16 model are removed, and the remaining layers are used to extract features from the images.

Feature Extraction

The feature_extractor function extracts features from an image using the VGG16 model. The features are flattened and used to compute similarity with precomputed features of celebrity images.

Setup

1. Clone the repository:

   git clone https://github.com/Aditya-Ranjan1234/BollywoodCeleb_Classifier.git
   cd BollywoodCeleb_Classifier

2. Create a virtual environment:

   python -m venv venv

3. Activate the virtual environment:

   • On Windows:

     venv\Scripts\activate

   • On macOS/Linux:

     source venv/bin/activate

4. Install the required packages:

   pip install -r requirements.txt

5. Download the pre-trained VGG16 model weights: The weights are automatically downloaded when running the app for the first time.

Usage

IMP : First run featureExtractor.ipynb to get features.pkl

1. Run the Streamlit app:

   streamlit run app.py

2. Upload an image:

   • Go to the web browser where the Streamlit app is running.
   • Upload an image of a face.

3. View the results:

   • The app will display the uploaded image and the most similar Bollywood celebrity image along with the predicted name.

License

This project is licensed under the GNU General Public License v3.0. See the LICENSE file for details.