CoReD: Generalizing Fake Media Detection with Continual Representation using Distillation (ACMMM'21 Oral Paper)

Overview

CoReD: Generalizing Fake Media Detection with Continual Representation using Distillation (ACMMM'21 Oral Paper)

(Accepted for oral presentation at ACMMM '21)

Paper Link: (arXiv) (ACMMM version)

CLRNet-pipeline

CLRNet-pipeline

Overview

We propose Continual Representation using Distillation (CoReD) method that employs the concept of Continual Learning (CL), Representation Learning (RL), and Knowledge Distillation (KD).

Comparison Baselines

  • Transfer-Learning (TL) : The first method is Transfer learning, where we perform fine-tuning on the model to learning the new Task.
  • Distillaion Loss (DL) : The third method is a part of our ablation study, wherewe only use the distillation loss component from our CoReD loss function to perform incremental learning.
  • Transferable GAN-generated Images Detection Framewor (TG) : The second method is a KD-based GAN image detection framework using L2-SP and self-training.

Requirements and Installation

We recommend the installation using the requilrements.txt contained in this Github.

python==3.8.0
torchvision==0.9.1
torch==1.8.1
sklearn
numpy
opencv_python

pip install -r requirements.txt

- Train & Evaluation

- Full Usages

  -m                   Model name = ['CoReD','KD','TG','FT']
  -te                  Turn on test mode True/False
  -s                   Name of 'Source' datasets. one or multiple names. (ex. DeepFake / DeepFake_Face2Face / DeepFake_Face2Face_FaceSwap)
  -t                   Name of 'Target' dataset. only a single name. (ex.DeepFake / Face2Face / FaceSwap / NeuralTextures) / used for Train only')
  -folder1             Sub-name of folder in Save path when model save
  -folder2             'name of folder that will be made in folder1 (just option)'
  -d                   Folder of path must contains Sources & Target folder names
  -w                   You can select the full path or folder path included in the '.pth' file
  -lr                  Learning late (For training)
  -a                   Alpha of KD-Loss
  -nc                  Number of Classes
  -ns                  Number of Stores
  -me                  Number of Epoch (For training)
  -nb                  Batch-Size
  -ng                  GPU-device can be set as ei 0,1,2 for multi-GPU (default=0) 

- Train

To train and evaluate the model(s) in the paper, run this command:

  • Task1 We must train pre-trained single model for task1 .
    python main.py -s={Source Name} -d={folder_path} -w={weights}  
    python main.py -s=DeepFake -d=./mydrive/dataset/' #Example 
    
  • Task2 - 4
    python main.py -s={Source Name} -t={Target Name} -d={folder_path} -w={weights}  
    python main.py -s=Face2Face_DeepFake -t=FaceSwap -d=./mydrive/dataset/ -w=./weights' #Example
    
  • Note that If you set -s=Face2Face_DeepFake -t=FaceSwap -d=./mydrive/dataset -w=./weights when you start training, data path "./mydrive/dataset" must include 'Face2Face', 'DeepFake', and 'FaceSwap', and these must be contained the 'train','val' folder which include 'real'&'fake' folders.

- Evaluation

After train the model, you can evaluate the dataset.

  • Eval
    python main.py -d= -w={weights} --test  
    python main.py -d=./mydrive/dataset/DeepFake/testset -w=./weights/bestmodel.pth --test #Example
    

- Result

  • AUC scores (%) of various methods on compared datasets.

- Task1 (GAN datasets and FaceForensics++ datasets)

- Task2 - 4

Citation

If you find our work useful for your research, please consider citing the following papers :)

@misc{kim2021cored,
    title={CoReD: Generalizing Fake Media Detection with Continual Representation using Distillation},
    author={Minha Kim and Shahroz Tariq and Simon S. Woo},
    year={2021},
    eprint={2107.02408},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}

- Contect

If you have any questions, please contact us at kimminha/[email protected]

- License

The code is released under the MIT license. Copyright (c) 2021

Owner
Minha Kim
@DASH-Lab on Sungkyunkwan University in Korea
Minha Kim
Imposter-detector-2022 - HackED 2022 Team 3IQ - 2022 Imposter Detector

HackED 2022 Team 3IQ - 2022 Imposter Detector By Aneeljyot Alagh, Curtis Kan, Jo

Joshua Ji 3 Aug 20, 2022
Strongly local p-norm-cut algorithms for semi-supervised learning and local graph clustering

Strongly local p-norm-cut algorithms for semi-supervised learning and local graph clustering

Meng Liu 2 Jul 19, 2022
This is a repository with the code for the ACL 2019 paper

The Story of Heads This is the official repo for the following papers: (ACL 2019) Analyzing Multi-Head Self-Attention: Specialized Heads Do the Heavy

231 Nov 15, 2022
The code for paper "Learning Implicit Fields for Generative Shape Modeling".

implicit-decoder The tensorflow code for paper "Learning Implicit Fields for Generative Shape Modeling", Zhiqin Chen, Hao (Richard) Zhang. Project pag

Zhiqin Chen 353 Dec 30, 2022
Liecasadi - liecasadi implements Lie groups operation written in CasADi

liecasadi liecasadi implements Lie groups operation written in CasADi, mainly di

Artificial and Mechanical Intelligence 14 Nov 05, 2022
PyTorch implementation of deep GRAph Contrastive rEpresentation learning (GRACE).

GRACE The official PyTorch implementation of deep GRAph Contrastive rEpresentation learning (GRACE). For a thorough resource collection of self-superv

Big Data and Multi-modal Computing Group, CRIPAC 186 Dec 27, 2022
Official implementation of ETH-XGaze dataset baseline

ETH-XGaze baseline Official implementation of ETH-XGaze dataset baseline. ETH-XGaze dataset ETH-XGaze dataset is a gaze estimation dataset consisting

Xucong Zhang 134 Jan 03, 2023
“英特尔创新大师杯”深度学习挑战赛 赛道3:CCKS2021中文NLP地址相关性任务

基于 bert4keras 的一个baseline 不作任何 数据trick 单模 线上 最高可到 0.7891 # 基础 版 train.py 0.7769 # transformer 各层 cls concat 明神的trick https://xv44586.git

孙永松 7 Dec 28, 2021
Code accompanying paper: Meta-Learning to Improve Pre-Training

Meta-Learning to Improve Pre-Training This folder contains code to run experiments in the paper Meta-Learning to Improve Pre-Training, NeurIPS 2021. P

28 Dec 31, 2022
LiDAR R-CNN: An Efficient and Universal 3D Object Detector

LiDAR R-CNN: An Efficient and Universal 3D Object Detector Introduction This is the official code of LiDAR R-CNN: An Efficient and Universal 3D Object

TuSimple 295 Jan 05, 2023
SpineAI Bilsky Grading With Python

SpineAI-Bilsky-Grading SpineAI Paper with Code 📫 Contact Address correspondence to J.T.P.D.H. (e-mail: james_hallinan AT nuhs.edu.sg) Disclaimer This

<a href=[email protected]"> 2 Dec 16, 2021
Tools to create pixel-wise object masks, bounding box labels (2D and 3D) and 3D object model (PLY triangle mesh) for object sequences filmed with an RGB-D camera.

Tools to create pixel-wise object masks, bounding box labels (2D and 3D) and 3D object model (PLY triangle mesh) for object sequences filmed with an RGB-D camera. This project prepares training and t

305 Dec 16, 2022
SASM - simple crossplatform IDE for NASM, MASM, GAS and FASM assembly languages

SASM (SimpleASM) - простая кроссплатформенная среда разработки для языков ассемблера NASM, MASM, GAS, FASM с подсветкой синтаксиса и отладчиком. В SA

Dmitriy Manushin 5.6k Jan 06, 2023
this is a lite easy to use virtual keyboard project for anyone to use

virtual_Keyboard this is a lite easy to use virtual keyboard project for anyone to use motivation I made this for this year's recruitment for RobEn AA

Mohamed Emad 3 Oct 23, 2021
dyld_shared_cache processing / Single-Image loading for BinaryNinja

Dyld Shared Cache Parser Author: cynder (kat) Dyld Shared Cache Support for BinaryNinja Without any of the fuss of requiring manually loading several

cynder 76 Dec 28, 2022
Text-Based Ideal Points

Text-Based Ideal Points Source code for the paper: Text-Based Ideal Points by Keyon Vafa, Suresh Naidu, and David Blei (ACL 2020). Update (June 29, 20

Keyon Vafa 37 Oct 09, 2022
Implementation of a protein autoregressive language model, but with autoregressive infilling objective (editing subsequences capability)

Protein GLM (wip) Implementation of a protein autoregressive language model, but with autoregressive infilling objective (editing subsequences capabil

Phil Wang 17 May 06, 2022
Research code for CVPR 2021 paper "End-to-End Human Pose and Mesh Reconstruction with Transformers"

MeshTransformer ✨ This is our research code of End-to-End Human Pose and Mesh Reconstruction with Transformers. MEsh TRansfOrmer is a simple yet effec

Microsoft 473 Dec 31, 2022
A3C LSTM Atari with Pytorch plus A3G design

NEWLY ADDED A3G A NEW GPU/CPU ARCHITECTURE OF A3C FOR SUBSTANTIALLY ACCELERATED TRAINING!! RL A3C Pytorch NEWLY ADDED A3G!! New implementation of A3C

David Griffis 532 Jan 02, 2023
HAT: Hierarchical Aggregation Transformers for Person Re-identification

HAT: Hierarchical Aggregation Transformers for Person Re-identification

11 Sep 05, 2022