Deepfake Network Architecture Attribution

Authors: Tianyun Yang, Ziyao Huang, Juan Cao, Lei Li, Xirong Li4662-4670

AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on multiple cross-test setups and a large-scale dataset demonstrate the effectiveness of DNA-Det.
Researcher Affiliation Academia 1 Key Lab of Intelligent Information Processing, Institute of Computing Technology, CAS, Beijing, China 2 University of Chinese Academy of Sciences, Beijing, China 3 Key Lab of Data Engineering and Knowledge Engineering, Renmin University of China
Pseudocode No The paper describes the proposed method using text and figures (Figure 4, Figure 6), but it does not include any formal pseudocode or algorithm blocks.
Open Source Code No The paper does not contain any explicit statement about making its source code publicly available, nor does it provide any links to a code repository.
Open Datasets Yes all of which are trained on celeb A dataset (Liu et al. 2015). We apply these transformations on a natural image dataset containing LSUN (Yu et al. 2015) and Celeb A.
Dataset Splits No The paper describes various 'cross-test setups' (cross-seed, cross-loss, cross-finetune, cross-dataset) and mentions 'train-set' in Table 1 for defining experiment conditions. However, it does not provide specific details about standard training, validation, and test splits (e.g., percentages or exact counts) for reproducibility.
Hardware Specification No The paper describes the network architecture (e.g., shallow 8-layer CNN) and image processing steps, but it does not specify any hardware details such as GPU models, CPU types, or memory used for conducting the experiments.
Software Dependencies No The paper mentions using 'Adam optimizer' but does not specify any software dependencies (e.g., libraries, frameworks) with their version numbers.
Experiment Setup Yes For optimization, we choose Adam optimizer. For the celeb A experiment in section , the initial learning rate is set to 10^-4 and is multiplied by 0.9 for every 500 iterations. For the LSUN-bedroom experiment in section and the experiment in section , the initial learning rate is set to 10^-3 and is multiplied by 0.9 for every 2500 iterations. The batch size is 32 #classes in Section and 16 #classes in Section with a class balance strategy.