This is the official repository of Video Face Manipulation Detection Through Ensemble of CNNs, submitted to ICPR2020 and currently available on arXiv.
We participate as ISPL team on Kaggle Deepfake Detection Challenge. With this implementation, we reached the 43rd position over 2116 teams (top 2%) on the private leaderboard.
This repository is currently under mantainance, feel free to notify us any lack by opening an issue.
- Install conda
- Create the
icpr2020
environment with environment.yml
$ conda env create -f environment.yml
$ conda activate icpr2020
You need to preprocess dataset in order to index all the samples and extract faces. Just run the script make_dataset.sh
$ ./scripts/make_dataset.sh
Please notice that we use only 32 frames per video. You can tweak easily tweak this parameter in extract_faces.py
In train_all.sh you can find a comprehensive list of all the commands for training the models presented in the paper. Please refer to the comments into the script for hints on their usage.
If you want to train some models without referring to the script:
- for the non-siamese architectures (e.g. EfficientNetB4, EfficientNetB4Att), you can simply specify the model in train_binclass.py as the --net parameter;
- for the siamese architectures (e.g. EfficientNetB4ST, EfficientNetB4AttST), you have to:
- train the architecture as a feature extractor first, using the train_triplet.py script and being careful of specifying its name in the --net parameter without the ST suffix. For instance, for training the EfficientNetB4ST you will have to first run
python train_triplet.py --net EfficientNetB4 --otherparams
; - finetune the model using train_binclass.py, being careful this time to specify the architecture's name with the ST suffix and to insert as the --init argument the path to the weights of the feature extractor trained at the previous step. You will end up running something like
python train_binclass.py --net EfficientNetB4ST --init path/to/EfficientNetB4/weights/trained/with/train_triplet/weights.pth --otherparams
- train the architecture as a feature extractor first, using the train_triplet.py script and being careful of specifying its name in the --net parameter without the ST suffix. For instance, for training the EfficientNetB4ST you will have to first run
In test_all.sh you can find a comprehensive list of all the commands for testing the models presented in the paper.
We also provide pretrained weights for all the architectures presented in the paper.
Please refer to this Dropbox link.
Each directory is named $NETWORK_$DATASET
where $NETWORK
is the architecture name and $DATASET
is the training dataset.
In each directory, you can find bestval.pth
which are the best network weights according to the validation set.
Additionally, you can find notebooks for results computations in the notebook folder.
Image and Sound Processing Lab - Politecnico di Milano
- Nicolò Bonettini ([email protected])
- Edoardo Daniele Cannas ([email protected])
- Sara Mandelli ([email protected])
- Luca Bondi ([email protected])
- Paolo Bestagini ([email protected])