FEditNet: Few-Shot Editing of Latent Semantics in GAN Spaces
Authors: Mengfei Xia, Yezhi Shu, Yuji Wang, Yu-Kun Lai, Qiang Li, Pengfei Wan, Zhongyuan Wang, Yong-Jin Liu
AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Qualitative and quantitative results demonstrate that our method outperforms the state-of-the-art methods on various datasets. The code is available at https://github.com/THU-LYJ-Lab/FEdit Net. ... We evaluated FEdit Net on: facial datasets Celeb A dataset (Liu et al. 2015), FFHQ dataset (Karras, Laine, and Aila 2019) and Danbooru2018 dataset (Anonymous, community, and Branwen 2021), animal dataset AFHQ dataset (Choi et al. 2020), scene datasets LSUN dataset (Yu et al. 2015) including church, tower and car. ... Implementation details. We trained FEdit Net on the platform of PyTorch (Paszke et al. 2019), in a Linux environment with an Nvidia A100 PCIe GPU. |
| Researcher Affiliation | Collaboration | 1Department of Computer Science, BNRist, Tsinghua University 2School of Computer Science and Informatics, Cardiff University 3Kuaishou Technology |
| Pseudocode | No | The paper does not contain any sections explicitly labeled 'Pseudocode' or 'Algorithm', nor are there any structured algorithm blocks. |
| Open Source Code | Yes | The code is available at https://github.com/THU-LYJ-Lab/FEdit Net. |
| Open Datasets | Yes | Datasets. We evaluated FEdit Net on: facial datasets Celeb A dataset (Liu et al. 2015), FFHQ dataset (Karras, Laine, and Aila 2019) and Danbooru2018 dataset (Anonymous, community, and Branwen 2021), animal dataset AFHQ dataset (Choi et al. 2020), scene datasets LSUN dataset (Yu et al. 2015) including church, tower and car. |
| Dataset Splits | No | In each dataset, we manually select 30 synthesized images containing the target attribute as our training dataset. The paper mentions a training dataset size but does not specify a separate validation split or explicit training/validation/test splits of the larger datasets used. |
| Hardware Specification | Yes | We trained FEdit Net on the platform of PyTorch (Paszke et al. 2019), in a Linux environment with an Nvidia A100 PCIe GPU. |
| Software Dependencies | No | We trained FEdit Net on the platform of PyTorch (Paszke et al. 2019). While PyTorch is mentioned, a specific version number is not provided, which is required for reproducibility of ancillary software. |
| Experiment Setup | Yes | The whole 30,000 training steps were completed in 6 hours to obtain editing directions of the best quality. ... we use the Adam (Nothaft et al. 2015) optimizer to simutaneously optimize the direction generator Gedit( ) and attribute assessor Dattr( ). The learning rate of Gedit( ) is set to 2e-3 while that of Dattr( ) is set to 2e-4. We use a fixed length of l = 5 in Eq. (3). |