detector developed for the AA2 course. This multimodel project allows users to distinguish between fake 🐛 and real pictures of faces. Based on CNN and transfer learning ⏩ properties.
Fake face detector developed for the AA2 course. Multimodel project designed for detecting fake face images from different people, angles and perspectives. Developed in python through the pytorch framework, it takes advange of Convolutional Neural Networks (CNN) properties for performing an accurate classification. This repository contains the arquitecture of two different models: a "made from scratch" model and a tranfer learning model generated with the convolutional layers of the well known model resNet18.
FakeFaceAuthenticity: a completely new architecture, designed from zero, with two convolutional layers, two pooling layers and two fully connected layers.
Transfer_learning: transfer learning model based on resNet18, two fully connected layers.
Activation Function: RELU
Optimization method: ADAMW (adaptative)
Loss Function: Cross Entropy
(c) 2024 Pablo Guilló
GitHub: Personal Repository
(c) 2024 Javier Franco
Github: https://github.com/Javierfg2