Independent Feature Decomposition and Instance Alignment for Unsupervised Domain Adaptation
Authors: Qichen He, Siying Xiao, Mao Ye, Xiatian Zhu, Ferrante Neri, Dongde Hou
IJCAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiment results demonstrate that our method achieves state-of-the-art performance on popular UDA benchmarks. The appendix and code are available at https://github.com/ayombeach/Ind UDA. ... 4 Experiments 4.1 Experimental Setup Datasets. We conduct experiments on three standard domain adaptation datasets. Office-31 [Saenko et al., 2010] is a popular dataset with a total of 4652 photos, including 31 object categories in three domains: Amazon (A), DSLR (D) and Webcam (W). Office-Home [Venkateswara et al., 2017] is a more challenging dataset, which contains 15500 images with an average of 70 images per class. It consists of 65 categories in 4 domains: Art (A), Clipart (C), Product (P) and Real World (R). Vis DA-2017 [Peng et al., 2018] is a simulation-to-real dataset across 12 categories. ... Tables 1-2 show experiment results of our Ind UDA and other state-of-the-art methods on datasets Office-home, Visda-2017, respectively. |
| Researcher Affiliation | Academia | Qichen He1 , Siying Xiao1 , Mao Ye1 , Xiatian Zhu2 , Ferrante Neri3 and Dongde Hou4 1School of CSE, University of Electronic Science and Technology of China, Chengdu, China 2Surrey Institute for People-Centred Artificial Intelligence, CVSSP, University of Surrey, Guildford, UK 3NICE Research Group, Department of Computer Science, University of Surrey, Guildford, UK 4Advanced Research Institute, Southwest University of Political Science&Law, Chongqing, China {202121081209, 2018270101016}@std.uestc.edu.cn, maoye@uestc.edu.cn, {xiatian.zhu, f.neri}@surrey.ac.uk, gxin 001@163.com |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | The appendix and code are available at https://github.com/ayombeach/Ind UDA. |
| Open Datasets | Yes | We conduct experiments on three standard domain adaptation datasets. Office-31 [Saenko et al., 2010] is a popular dataset with a total of 4652 photos, including 31 object categories in three domains: Amazon (A), DSLR (D) and Webcam (W). Office-Home [Venkateswara et al., 2017] is a more challenging dataset, which contains 15500 images with an average of 70 images per class. It consists of 65 categories in 4 domains: Art (A), Clipart (C), Product (P) and Real World (R). Vis DA-2017 [Peng et al., 2018] is a simulation-to-real dataset across 12 categories. |
| Dataset Splits | No | Vis DA-2017 [Peng et al., 2018] is a simulation-to-real dataset across 12 categories. The source domain training set contains 152397 synthetic images and the target domain validation set contains 55388 real-world images. |
| Hardware Specification | No | No specific hardware specifications (e.g., GPU/CPU models, memory) are mentioned for the experimental setup. |
| Software Dependencies | No | All experiments are carried out on the Pytorch platform. |
| Experiment Setup | Yes | The invertible flow INN consists of 5 blocks, whose structure is shown in Appendix, and the subnet ϕ1,2,3( ) of each INN block consists of 3 convolution layers with Leaky Re Lu function only in the first layer. In order to compare the state-of-the-art methods fairly across various datasets, we use the commonly pre-trained Res Net-50/50/101 [He et al., 2016a; He et al., 2016b] on Image Net [Deng et al., 2009] as feature extractors for Office31, Office-Home and Vis DA-2017 respectively. The classifier consists of a fully connected layer. With the exception of the batch normalization layers domain-specific characteristics, all network parameters are shared by the source domain and target domain data. The initial learning rate η0 is set as 0.001 for the first convolutional layers and 0.01 for the rest including the entire INN network and no pre-trained layers in the feature extractor. We adopt the same learning rate scheduler ηp = η0(1 + ap) b. For Office-31 and Office-home datasets, we set a = 10 and b = 0.75, while for Visda-2017 we set a = 10 and b = 2.25. We select half of the channels as domain-invariant feature channels in channel mask block and choose 0.5, 0.3 as the value of γ and λ respectively. |