Capsule-Forensics

Implementation of the paper: Capsule-Forensics: Using Capsule Networks to Detect Forged Images and Videos (ICASSP 2019).

UPDATE: The Capsule-Forensics-v2 has been released here, corresponding to this extended work.

You can clone this repository into your favorite directory:

$ git clone https://github.com/nii-yamagishilab/Capsule-Forensics

Requirement

PyTorch 0.3
TorchVision
scikit-learn
Numpy
pickle

Project organization

Datasets folder, where you can place your training, evaluation, and test set:
```
./dataset/<train; validation; test>
```
Checkpoint folder, where the training outputs will be stored:
```
./checkpoints
```

Pre-trained models for Face2Face and DeepFakes detection (with settings described in our paper) are provided in the checkpoints folder.

Dataset

Each dataset has two parts:

Real images: ./dataset/<train;test;validation>/Real
Fake images: ./dataset/<train;test;validation>/Fake

The dataset need to be pre-processed to crop facial area.

Training

Note: Parameters with detail explanation could be found in the corresponding source code.

$ python train.py --dataset dataset --train_set train --val_set validation --outf checkpoints --batchSize 100 --niter 100

Evaluating

Note: Parameters with detail explanation could be found in the corresponding source code.

For testing on image level, using test.py

$ python test.py --dataset dataset --test_set test --outf checkpoints --id <your selected id>

For testing on large images using patch aggregation strategy, please use test_by_patches.py.

For testing on video level by aggregating the predicted probabilities of video frames, please use test_vid_lvl.py

Authors

Huy H. Nguyen (https://researchmap.jp/nhhuy/?lang=english)
Junichi Yamagishi (https://researchmap.jp/read0205283/?lang=english)
Isao Echizen (https://researchmap.jp/echizenisao/?lang=english)

Acknowledgement

This work was supported by JSPS KAKENHI Grant Numbers (16H06302, 17H04687, 18H04120, 18H04112, 18KT0051) and by JST CREST Grant Number JPMJCR18A6, Japan.

Reference

H. H. Nguyen, J. Yamagishi, and I. Echizen, “Capsule-Forensics: Using Capsule Networks to Detect Forged Images and Videos,” Proc. of the 2019 International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2019), 5 pages, (May 2019)