Exact dataset labels
yoctta opened this issue · comments
Thank you for releasing the dataset, it is really valuable for deepfake researches. Could you release the labels of testset and the exact manipulation methods of 16cls_labels?
1: FaceShifter
2: FS-GAN
3: DeepFakes
4: BlendFace
5: MMReplacement
6: DeepFakes-StarGAN-Stack
7: Talking Head Video
8: ATVG-Net
9: StarGAN-BlendFace-Stack
10: First Order Motion
11: StyleGAN2
12: MaskGAN
13: StarGAN2
14: SC-FEGAN
15: DiscoFaceGAN
The label of test set will be public soon.
1: FaceShifter 2: FS-GAN 3: DeepFakes 4: BlendFace 5: MMReplacement 6: DeepFakes-StarGAN-Stack 7: Talking Head Video 8: ATVG-Net 9: StarGAN-BlendFace-Stack 10: First Order Motion 11: StyleGAN2 12: MaskGAN 13: StarGAN2 14: SC-FEGAN 15: DiscoFaceGAN
The label of test set will be public soon.
Hi, where can i find the label of test set? Have them already been released?
1: FaceShifter 2: FS-GAN 3: DeepFakes 4: BlendFace 5: MMReplacement 6: DeepFakes-StarGAN-Stack 7: Talking Head Video 8: ATVG-Net 9: StarGAN-BlendFace-Stack 10: First Order Motion 11: StyleGAN2 12: MaskGAN 13: StarGAN2 14: SC-FEGAN 15: DiscoFaceGAN
The label of test set will be public soon.
What are labels of 16~19？
1: FaceShifter 2: FS-GAN 3: DeepFakes 4: BlendFace 5: MMReplacement 6: DeepFakes-StarGAN-Stack 7: Talking Head Video 8: ATVG-Net 9: StarGAN-BlendFace-Stack 10: First Order Motion 11: StyleGAN2 12: MaskGAN 13: StarGAN2 14: SC-FEGAN 15: DiscoFaceGAN
The label of test set will be public soon.What are labels of 16~19？
real image
You mentioned four real sources in the paper. Actually, I would like to know which source each number corresponds to. Thx!
16: RAVDESS
17: CREMA-D
18: AVSpeech
19: vox
Are the test set labels available now?
1: FaceShifter 2: FS-GAN 3: DeepFakes 4: BlendFace 5: MMReplacement 6: DeepFakes-StarGAN-Stack 7: Talking Head Video 8: ATVG-Net 9: StarGAN-BlendFace-Stack 10: First Order Motion 11: StyleGAN2 12: MaskGAN 13: StarGAN2 14: SC-FEGAN 15: DiscoFaceGAN
The label of test set will be public soon.
In third row in the file 'image_list.txt':
train_release/7/2ac43aa43bf473f9a9c09b4b608619d3/b039e05887ddaad1a72964d5d1ab88d3/frame00038.jpg 1 1 15
The 16cls_label '15' is different from category '7'. So I should use which one to determine corresponding method? thx~
1: FaceShifter 2: FS-GAN 3: DeepFakes 4: BlendFace 5: MMReplacement 6: DeepFakes-StarGAN-Stack 7: Talking Head Video 8: ATVG-Net 9: StarGAN-BlendFace-Stack 10: First Order Motion 11: StyleGAN2 12: MaskGAN 13: StarGAN2 14: SC-FEGAN 15: DiscoFaceGAN
The label of test set will be public soon.In third row in the file 'image_list.txt': train_release/7/2ac43aa43bf473f9a9c09b4b608619d3/b039e05887ddaad1a72964d5d1ab88d3/frame00038.jpg 1 1 15 The 16cls_label '15' is different from category '7'. So I should use which one to determine corresponding method? thx~
7
Allow me to confirm the above conclusions.
train_release/7/2ac43aa43bf473f9a9c09b4b608619d3/b039e05887ddaad1a72964d5d1ab88d3/frame00038.jpg 1 1 15for example- 7
- the subdirectory number == the method index? 🤔
- 15
- What do the n-way numbers in
image_list.txtmean? Can this be ignored?🤔
- What do the n-way numbers in
- 7
Counting results for subdirectory numbers and labels in the image_list.txt
Is this relationship between sub-dir and method correct? 🤔
sub-dir two-way three-way n-way count method
1 1 2 8 99687 FaceShifter
2 1 1 7 199993 FS-GAN
3 1 2 3 100000 DeepFakes
4 1 2 2 100000 BlendFace
5 1 2 1 100000 MMReplacement
6 1 2 4 100000 DeepFakes-StarGAN-Stack
7 1 1 15 21570 Talking Head Video < train_release/7/2ac43aa43bf473f9a9c09b4b608619d3/b039e05887ddaad1a72964d5d1ab88d3/frame00038.jpg 1 1 15
8 1 1 10 96880 ATVG-Net
9 1 1 12 53947 StarGAN-BlendFace-Stack
10 1 1 14 29467 First Order Motion
11 1 1 13 29570 StyleGAN2
12 1 2 9 14785 MaskGAN
13 1 2 6 99949 StarGAN2
14 1 1 11 58444 SC-FEGAN
15 1 2 5 88089 DiscoFaceGAN
16 0 0 0 57047 RAVDESS
17 0 0 0 73736 CREMA-D
18 0 0 0 553264 AVSpeech
19 0 0 0 474877 VoxCeleb2Yes, thank you for your example, this explanation is completely correct!