CDTrans: Cross-domain Transformer for Unsupervised Domain Adaptation

Authors: Tongkun Xu, Weihua Chen, Pichao WANG, Fan Wang, Hao Li, Rong Jin

ICLR 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments show that our proposed method achieves the best performance on public UDA datasets, e.g. Vis DA-2017 and Domain Net.
Researcher Affiliation Collaboration Tongkun Xu12 , Weihua Chen1 , Pichao Wang1, Fan Wang1, Hao Li1, Rong Jin1 1Alibaba Group, 2Shandong University
Pseudocode No No pseudocode or clearly labeled algorithm blocks are present in the paper.
Open Source Code Yes Code and models are available at https: //github.com/CDTrans/CDTrans.
Open Datasets Yes The proposed method is verified on four popular UDA benchmarks, including Vis DA-2017 (Peng et al., 2017), Office-Home (Venkateswara et al., 2017), Office-31 (Saenko et al., 2010) and Domain Net (Peng et al., 2019).
Dataset Splits No The paper mentions using well-known UDA benchmarks (Vis DA-2017, Office-Home, Office-31, Domain Net), which typically have predefined splits. However, it does not explicitly state the train/validation/test dataset splits (e.g., percentages or counts) within the paper's text.
Hardware Specification No No specific hardware details (e.g., GPU/CPU models, memory) used for running experiments are mentioned in the paper.
Software Dependencies No The paper mentions using Dei T-small and Dei T-base as backbones and Stochastic Gradient Descent, but does not provide specific version numbers for software dependencies like deep learning frameworks (e.g., PyTorch, TensorFlow) or programming languages.
Experiment Setup Yes The input image size in our experiments is 224 224. Both the Dei T-small and Dei T-base (Touvron et al., 2021) are adopted as our backbone for fair comparison. We use the Stochastic Gradient Descent algorithm with the momentum of 0.9 and weight decay ratio 1e-4 to optimize the training process. The learning rate is set to 3e-3 for Office-Home, Office-31 and Domain Net, 5e-5 for Vis DA-2017 since it can easily converge. The batch size is set to 64.