Cross-Domain Feature Augmentation for Domain Generalization
Authors: Yingnan Liu, Yingtian Zou, Rui Qiao, Fusheng Liu, Mong Li Lee, Wynne Hsu
IJCAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We implement our proposed solution using Py Torch 1.12.0 and perform a series of experiments on NVIDIA Tesla V100 GPU to evaluate the effectiveness of the proposed XDomain Mix. The following benchmark datasets are used: Camelyon17 [Bandi et al., 2018] from Wilds [Koh et al., 2021]. This dataset contains 455,954 tumor and normal tissue slide images from 5 hospitals (domains). ... Experiments on widely used benchmark datasets demonstrate that our proposed method is able to achieve state-of-the-art performance. Quantitative analysis indicates that our feature augmentation approach facilitates the learning of effective models that are invariant across different domains. |
| Researcher Affiliation | Academia | 1School of Computing, National University of Singapore 2Institute of Data Science, National University of Singapore |
| Pseudocode | No | The paper describes its proposed method in Section 3 using textual descriptions and mathematical equations (e.g., Equation 5, 6, 7), but it does not include a dedicated pseudocode block or algorithm listing. |
| Open Source Code | No | The paper does not provide a direct link to source code, nor does it explicitly state that the code for their method will be made publicly available. |
| Open Datasets | Yes | The following benchmark datasets are used: Camelyon17 [Bandi et al., 2018] from Wilds [Koh et al., 2021]. This dataset contains 455,954 tumor and normal tissue slide images from 5 hospitals (domains). FMo W [Christie et al., 2018] from Wilds. This dataset contains 141,696 satellite images from 62 land use categories across 16 years from 5 regions (domains). PACS [Li et al., 2017]. This dataset contains 9,991 images of 7 objects in 4 visual styles (domains): art painting, cartoon, photo, and sketch. Terra Incognita [Beery et al., 2018]. The dataset contains 24,788 images from 10 categories of wild animals taken from 4 different locations (domains). Domain Net [Peng et al., 2019]. This dataset contains 586,575 images from 365 classes in 6 visual styles (domains): clipart, infograph, painting, quickdraw, real, and sketch. |
| Dataset Splits | Yes | For Camelyon17 and FMo W datasets, we follow the setup in LISA [Yao et al., 2022a]. Non-pretrained Dense Net-121 is used for Canmelyon17 and pretrained Dense Net-121 is used for FMo W. We use the same partitioning in Wilds [Koh et al., 2021] to obtain the training, validation, and test domains. The batch size is set to 32, and the model is trained for 2 epochs for Camelyon17 and 5 epochs for FMo W. The learning rate and weight decay are set to 1e-4 and 0. The warm-up phase is set to 4000 steps. We tune the step n in {100, 500} for changing τd. The best model is selected based on its performance in the validation domain. For PACS, Terra Incognita and Domain Net datasets, we follow the setup in Domain Bed [Gulrajani and Lopez-Paz, 2021], and use a pre-trained Res Net-50. Each domain in the dataset is used as a test domain in turn, with the remaining domains serving as training domains. The batch size is set to 32 (24 for Domain Net), and the model is trained for 5000 steps (15000 steps for Domain Net). We tune the learning rate in {2e-5, 3e-5, 4e-5, 5e-5, 6e-5} and weight decay in (1e-6, 1e-2) using the Domain Bed framework. The warm-up phase is set to 3000 steps and n is set to 100 steps. The best model is selected based on its performance on the validations split of the training domains. |
| Hardware Specification | Yes | We implement our proposed solution using Py Torch 1.12.0 and perform a series of experiments on NVIDIA Tesla V100 GPU to evaluate the effectiveness of the proposed XDomain Mix. |
| Software Dependencies | Yes | We implement our proposed solution using Py Torch 1.12.0 and perform a series of experiments on NVIDIA Tesla V100 GPU to evaluate the effectiveness of the proposed XDomain Mix. |
| Experiment Setup | Yes | The batch size is set to 32, and the model is trained for 2 epochs for Camelyon17 and 5 epochs for FMo W. The learning rate and weight decay are set to 1e-4 and 0. The warm-up phase is set to 4000 steps. We tune the step n in {100, 500} for changing τd. The batch size is set to 32 (24 for Domain Net), and the model is trained for 5000 steps (15000 steps for Domain Net). We tune the learning rate in {2e-5, 3e-5, 4e-5, 5e-5, 6e-5} and weight decay in (1e-6, 1e-2) using the Domain Bed framework. The warm-up phase is set to 3000 steps and n is set to 100 steps. |