Chimera

Setup and Installation

First create a conda environment using the following command:

conda env create -f environment.yml

Then activate the environment using the following command:

conda activate chimera

Then download the required data and models from this link

Test Recapture Detection

In the recapture-detection directory, run the following command:

python main.py --test --test_path TEST_PATH --config CONFIG --test_raw_dirnames RAW_DIRNAMES --test_recap_dirnames RECAP_DIRNAMES

where TEST_PATH is the path to the model, CONFIG is the path to the configuration file, RAW_DIRNAMES is the list of raw directory names, and RECAP_DIRNAMES is the list of recapture directory names. Ensure that the configuration file matches the model being tested. Results should be printed to the terminal.

Test Deepfake Detection

In the deepfake-detection directory, first update dataset paths in dataset_paths.py and move the deepfake detection model weights are in the pretrained_weights folder.

The following command:

./test.sh

will run all the deepfake detection tests at once and save the results in the deepfake-detection/results directory.

Training Chimera

To train Chimera, you must collect data using a fixed screen and camera setup. Then, use the pix2pix training script to train the model. Training is done in the pytorch-CycleGAN-and-pix2pix directory.

Examples of training commands are provided in the pytorch-CycleGAN-and-pix2pix/scripts/train_pix2pix.sh file; ideal parameters depend on setup. Refer to the paper for more details on the training process.