Fake vs Real Image Classification (Hackathon)

Description

This project is designed to classify images as either real or fake, specifically targeting altered images in car technical inspections. The model utilizes ResNet to differentiate between genuine and manipulated images.

Dataset

[The dataset](https://www.kaggle.com/datasets/adilakimshe/datasaur/ - dataset.) consists of images of cars, where some have been altered to manipulate inspection results. The images are preprocessed and augmented before being fed into the model.

Model Architecture

The model is based on ResNet, a deep convolutional neural network architecture known for its residual learning framework, making it highly effective for image classification tasks.

Installation & Dependencies

To run this project, install the required dependencies:

pip install tensorflow numpy pandas matplotlib seaborn opencv-python

Usage

Run the following command to train the model:

python train.py

To test the model on new images:

python predict.py --image path/to/image.jpg

Results

The model achieves high accuracy in distinguishing between real and fake images, making it a reliable solution for detecting fraudulent alterations in car inspections.

Acknowledgments

This project was developed as part of a hackathon challenge focused on image authenticity verification.