erprogs / GenConViT

Deepfake Video Detection Using Generative Convolutional Vision Transformer
GNU General Public License v3.0
51 stars 11 forks source link

Dataset Preprocess #11

Closed gamesdsad closed 4 months ago

gamesdsad commented 4 months ago

First of all thank you for sharing your project. What happens to data preprocessing when I want to train a new model with FF++ and DFDC data?

erprogs commented 4 months ago

Hi, extract the images from the dataset, and store it in sample train data format.

gamesdsad commented 4 months ago

Hi, extract the images from the dataset, and store it in sample train data format.

Thank you for your quick reply, I have a general understanding of the principle, you are referring to the FF++ and DFDC preprocessing, detection of face extraction frame, and frame data is divided into the sample_train_data format?

erprogs commented 4 months ago

For training the model, yes, during inference, you don't need that format. You just extract face regions from video frames and pass them to the model.

gamesdsad commented 4 months ago

I know what your mean.Thank you very much!!!!

For training the model, yes, during inference, you don't need that format. You just extract face regions from video frames and pass them to the model.