search

davide-coccomini / Combining-EfficientNet-and-Vision-Transformers-for-Video-Deepfake-Detection

Code for Video Deepfake Detection model from "Combining EfficientNet and Vision Transformers for Video Deepfake Detection" presented at ICIAP 2021.
https://dl.acm.org/doi/abs/10.1007/978-3-031-06433-3_19
MIT License
224 stars 57 forks source link

ff++dataset structure issue #55

Open Mrmao233 opened 1 month ago

Mrmao233 commented 1 month ago

The preprocessing process seems to have some problems with the ff++ dataset. The structure of the ff++ dataset I downloaded from the official script is as follows, but there will be various problems following the steps in readme. （I wanted to test the faceswap subdataset first, so I put the other datasets in the NOTNOW folder first） 1 Next, I type the commandpython detect_faces.py --dataset FACEFORENSICS --data_path ../datasets/ff++.Then The script started extracting faces and saved the json file in the boxes folder.(Since I only used the sub-dataset, I only needed to extract 2000 video files in total, and the numbers of generated json files were also 2,000) After that, I used the following command to extract faces.python extract_crops.py --data_path ../datasets/ff++ --output_path ../train_dataset/ --dataset FACEFORENSICS In the start, the script reported the following error. image So I went to modify the utils.py file.I deleted the extra "c23", although I don't know if it was the right thing to do image Then the script starts to run, but it keeps running with the error shown below, and also produces the structure in my output folder (I don't know if this is normal, because it seems different from the structure in readme, do I have to build the structure in readme myself?) image I came to the output folder, and this number doesn't seem right. image

Mrmao233 commented 1 month ago

Your method has greatly helped my research, so I would be very grateful if you could spare some time to help me solve the problem!