https://github.com/perpendicularai/imageclassification

A deep-learning notebook to detect real or fake faces. Now comes with it's own PyPI module.
https://github.com/perpendicularai/imageclassification

convolutional-neural-networks deep-learning deep-neural-networks image-classification

Last synced: 8 days ago
JSON representation

A deep-learning notebook to detect real or fake faces. Now comes with it's own PyPI module.

• Host: GitHub
• URL: https://github.com/perpendicularai/imageclassification
• Owner: perpendicularai
• Created: 2024-01-03T08:30:45.000Z (about 1 year ago)
• Default Branch: main
• Last Pushed: 2024-02-13T01:59:05.000Z (11 months ago)
• Last Synced: 2024-12-23T14:34:25.786Z (17 days ago)
• Topics: convolutional-neural-networks, deep-learning, deep-neural-networks, image-classification
• Language: Jupyter Notebook
• Homepage: https://perpendicular.web.za
• Size: 886 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
![image](https://github.com/perpendicularai/imageclassification/assets/146530480/6938491a-8670-47a1-8cb8-aed6fcad5978)

# Binary Classification

This repo his been put together to showcase the ability of using deep-learning to train a binary classification model. The model in particular has been trained to detect between real and fake faces. This is ideal for locations that require a high degree of security, whether it be banks, airports, government buildings, schools to name a few. The dataset used to train the model is 4GB's in size. A link is provided below. The model was trained for 100 epochs on an Intel i5 CPU with 8GB's of RAM, and thus can be deployed on most systems. 

## How to :

* Download dataset - [fakeface_dataset](https://www.kaggle.com/datasets/xhlulu/140k-real-and-fake-faces)

*  Set train and test data paths in the notebook provided

* Once model has been trained, run the following command to save the model to be used for inference `model.save('MODEL_DIRECTORY_NAME')`. This will save all the model files to your current directory in the directory name you provided.

* A test image has been provided of someone that does not exist. The image was generated using a popular GAN. See [test_image](https://github.com/perpendicularai/imageclassification/tree/main/test_image/) directory for more.

## Alternatively :

If you would like to get started right away with inference on an image, see [PyPI_HOWTO.md](https://github.com/perpendicularai/imageclassification/blob/main/PyPI_HOWTO.md)

## Training stats :

* Accuracy -

![image](https://github.com/perpendicularai/imageclassification/assets/146530480/df14fb83-568b-478b-8394-2022b3409818)

* Loss -

![image](https://github.com/perpendicularai/imageclassification/assets/146530480/743fbaf9-58a8-49e9-befe-9e3f32507597)