AniFaceGAN: Animatable 3D-Aware Face Image Generation for Video Avatars
Authors: Yue Wu, Yu Deng, Jiaolong Yang, Fangyun Wei, Qifeng Chen, Xin Tong
NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments demonstrate our superior performance over prior works. Project page: https://yuewuhkust.github.io/Ani Face GAN/ |
| Researcher Affiliation | Collaboration | 1HKUST 2Tsinghua University 3Microsoft Research |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Project page: https://yuewuhkust.github.io/Ani Face GAN/ |
| Open Datasets | Yes | We train our method on the FFHQ [26] dataset3 which contains 70K face images. ... FFHQ is released under the Creative Commons BY-NC-SA 4.0 license; the human face images therein were published on Flickr by their authors under licenses that all allow free use for non-commercial purposes. |
| Dataset Splits | No | The paper mentions training on FFHQ but does not specify explicit train/validation/test splits with percentages or sample counts for reproduction. |
| Hardware Specification | Yes | We train our models on 8 Nvidia Tesla V100 GPUs with a batch size of 32 at the resolution of 128 128. |
| Software Dependencies | No | In our experiments, the Adam optimizer [29] with β1 = 0 and β2 = 0.9 is applied for training our model. The paper mentions software components but does not provide specific version numbers for libraries or frameworks used. |
| Experiment Setup | Yes | We set the learning rate to 2e 5 for the deformation network and the generative radiance manifolds, and 2e 4 for the discriminator. We train our models on 8 Nvidia Tesla V100 GPUs with a batch size of 32 at the resolution of 128 128. |