Realistic Face Reenactment via Self-Supervised Disentangling of Identity and Pose

Authors: Xianfang Zeng, Yusu Pan, Mengmeng Wang, Jiangning Zhang, Yong Liu12757-12764

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate our model on Vox Celeb1 and Ra FD dataset. Experiment results demonstrate the superior quality of reenacted images and the flexibility of transferring facial movements between identities.
Researcher Affiliation Academia Institute of Cyber-Systems and Control, Zhejiang University, China {zzlongjuanfeng, corenel, mengmengwang, 186368}@zju.edu.cn, yongliu@iipc.zju.edu.cn
Pseudocode No No pseudocode or algorithm block is explicitly labeled or presented in a structured format.
Open Source Code No No explicit statement about releasing source code or a link to a code repository is provided in the paper.
Open Datasets Yes We evaluate our model on Vox Celeb1 and Ra FD dataset. [...] In experiments, quantitative and qualitative comparisons are conducted on Vox Celeb1 and Ra FD dataset (Nagrani, Chung, and Zisserman 2017; Zhang et al. 2019).
Dataset Splits Yes We train all the models on the training and validation set and report their results on the corresponding test set. [...] Specifically, we randomly select 50 videos from the test set and 32 hold-out frames from each video. These frames are excluded from the fine-tuning process (if necessary) and used as driving images to be transformed from the remaining part in each video.
Hardware Specification Yes All experiments are conducted in a node with 2 NVIDIA RTX 2080Ti GPUs.
Software Dependencies No The paper mentions using the Adam optimizer but does not specify version numbers for any programming languages, libraries, or other software components used in the implementation.
Experiment Setup Yes The learning rate is set to 1 10 4, except for the discriminator, whose is 4 10 4. We use the Adam (Kingma and Ba 2015) optimizer with β1 = 0, β2 = 0.9 and decrease learning rate linearly. [...] We experimentally determine to only optimize the two embedders in the first 30 epochs.