Few-shot Hybrid Domain Adaptation of Image Generator
Authors: Hengjia Li, Yang Liu, Linxuan Xia, Yuqi Lin, Wenxiao Wang, Tu Zheng, Zheng Yang, Xiaohui Zhong, Xiaobo Ren, Xiaofei He
ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments show that our method can obtain numerous domain-specific attributes in a single adapted generator, which surpasses the baseline methods in semantic similarity, image fidelity, and cross-domain consistency. |
| Researcher Affiliation | Collaboration | Hengjia Li 1,4 Yang Liu 2 Linxuan Xia1 Yuqi Lin1 Wenxiao Wang 3 Tu Zheng 4 Zheng Yang4 Xiaohui Zhong5 Xiaobo Ren6 Xiaofei He1,4 1 State Key Lab of CAD&CG, Zhejiang University 2 Alibaba Cloud 3 Zhejiang University 4 Fabu Inc. 5 Ningbo Beilun Third Container Terminal Co., Ltd 6 Ningbo Zhoushan Port Co., Ltd |
| Pseudocode | No | The paper includes mathematical equations and diagrams (e.g., Figure 2) to describe the method, but it does not contain any clearly labeled pseudocode or algorithm blocks. |
| Open Source Code | No | The paper states: 'Project page is at https://echopluto.github.io/FHDA-project/.' This is a project overview page, not an explicit statement that the code is publicly available or a direct link to a code repository. |
| Open Datasets | Yes | Datasets: Following previous literature, we consider Flickr-Faces-HQ (FFHQ) (Karras et al., 2019) as one of the source domains... Swin (Liu et al., 2021) and Dinov2 (Oquab et al., 2023) as the image encoder, which are pre-trained on Image Net 22k dataset (Deng et al., 2009)... We adapt the pre-trained generator from LSUN Church (Yu et al., 2015)... |
| Dataset Splits | No | The paper mentions using '10 randomly sampled targets for each domain' and a 'training session' but does not specify explicit train/validation/test dataset splits with percentages, absolute counts, or methods like cross-validation to reproduce data partitioning. |
| Hardware Specification | Yes | Following the setting of previous DA methods, we utilize the batch size of 4 and a training session typically requires 300 iterations in roughly 3 minutes on a single NVIDIA TITAN GPU. |
| Software Dependencies | No | The paper mentions using 'Style GAN2' and 'Swin' and 'Dinov2' models, but it does not specify any software dependencies with explicit version numbers (e.g., Python, PyTorch, or CUDA versions) required for reproduction. |
| Experiment Setup | Yes | Following the setting of previous DA methods, we utilize the batch size of 4 and a training session typically requires 300 iterations in roughly 3 minutes on a single NVIDIA TITAN GPU. Besides, we set the balancing factor λ as 1 in Eq. (6) and Eq. (9). As depicted in Eq. (7) and Eq. (8), we use pre-defined domain coefficient to modulate the attributes from multiple domains. For most experiments on two domains, we use αi = 0.5 except for babysunglasses with αbaby = 0.3 and αsunglasses = 0.7. For the experiment of three domains, we set αbaby = 0.7, αsketch = 0.4, and αsmile = 0.3. |