CFFT-GAN: Cross-Domain Feature Fusion Transformer for Exemplar-Based Image Translation
Authors: Tianxiang Ma, Bingchuan Li, Wei Liu, Miao Hua, Jing Dong, Tieniu Tan
AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct rich quantitative and qualitative experiments on several image translation tasks, and the results demonstrate the superiority of our approach compared to state-of-the-art methods. Ablation studies show the importance of our proposed CFFT. |
| Researcher Affiliation | Collaboration | Tianxiang Ma1,2*, Bingchuan Li3, Wei Liu3, Miao Hua3, Jing Dong2 , Tieniu Tan2,4 1School of Artificial Intelligence, University of Chinese Academy of Sciences 2CRIPAC & NLPR, Institute of Automation, Chinese Academy of Sciences 3Byte Dance Ltd, Beijing, China 4Nanjing University |
| Pseudocode | No | The paper describes the model architecture and its components in prose and diagrams (Figure 2) but does not include a formal pseudocode or algorithm block. |
| Open Source Code | No | The paper does not provide an explicit statement about releasing source code or a link to a code repository. |
| Open Datasets | Yes | Datasets. To verify the generality of our approach, we perform experiments on several widely used public datasets. Deepfashion (Liu et al. 2016)... Celeb A-HQ (Liu et al. 2015)... Met Faces (Karras et al. 2020)... ADE20K (Zhou et al. 2017)... AFHQ (Choi et al. 2020)... |
| Dataset Splits | No | The paper lists several datasets used for experiments but does not provide specific details on how these datasets were split into training, validation, and test sets (e.g., exact percentages or sample counts for each split). |
| Hardware Specification | Yes | Our experiments are carried out on 8 32GB Tesla V100 GPUs. |
| Software Dependencies | No | The paper mentions using 'Adam solver' but does not provide specific version numbers for any software dependencies or libraries. |
| Experiment Setup | Yes | We use the TTUR (Heusel et al. 2017) learning strategy with a generator and discriminator learning rate of 1e-4 and 4e-4, respectively. We utilize Adam solver with β1 = 0 and β2 = 0.999. The input and output feature scales for the CFFT module are (C, H, W) = (64, 64, 64). The weights of the total loss are λalign = 10, λmatch = 10, λperc = 0.001, λCX = 10, λadv = 10. |