anonymous-author-sub/seeable

SeeABLE: Soft Discrepancies and Bounded Contrastive Learning for Exposing Deepfakes

SeeABLE: Soft Discrepancies and Bounded Contrastive Learning for Exposing Deepfakes

This is the codebase behind SeeABLE

Requirements

You need git and conda/mamba.

You need Hydra to run this framework See conf dir for possible config.

Setup

1. Clone this github repo and install its requirements:

git clone https://github.com/anonymous-author-sub/seeable.git
cd seeable

2. Mamba installation

If you don't have conda, you can install it yourself or choose mamba (faster). Follow this link to install it using mambaforge. Latest installers with Mamba in the base environment:

OS	Architecture	Download
Linux	x86_64 (amd64)	Mambaforge-Linux-x86_64
OS X	x86_64	Mambaforge-MacOSX-x86_64
OS X	arm64 (Apple Silicon)	Mambaforge-MacOSX-arm64
Windows	x86_64	Mambaforge-Windows-x86_64

After installing Mambaforge, verify that it is working by running the following command:

mamba --version

This should print the version of Mambaforge that you have installed.

3. Install the environment

The environment can be installed using conda/mamba

conda env create -f environment.yml

Instruction

Activate the environment

conda activate seeable

TODO

seeable's Issues

paper question?

Thx for this work! There are some questions during reading and I sincerely hope the authors can clear these confusions : )

After reading the paper, I still confused how to determine whether an image is real or fake... Does the final decision depends on anomaly score? Then for a testing image, should I need to modify the image 33 times to get the score ?