Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
119 tokens/sec
GPT-4o
56 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
6 tokens/sec
GPT-4.1 Pro
47 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Learning Generalizable Models via Disentangling Spurious and Enhancing Potential Correlations (2401.05752v1)

Published 11 Jan 2024 in cs.CV

Abstract: Domain generalization (DG) intends to train a model on multiple source domains to ensure that it can generalize well to an arbitrary unseen target domain. The acquisition of domain-invariant representations is pivotal for DG as they possess the ability to capture the inherent semantic information of the data, mitigate the influence of domain shift, and enhance the generalization capability of the model. Adopting multiple perspectives, such as the sample and the feature, proves to be effective. The sample perspective facilitates data augmentation through data manipulation techniques, whereas the feature perspective enables the extraction of meaningful generalization features. In this paper, we focus on improving the generalization ability of the model by compelling it to acquire domain-invariant representations from both the sample and feature perspectives by disentangling spurious correlations and enhancing potential correlations. 1) From the sample perspective, we develop a frequency restriction module, guiding the model to focus on the relevant correlations between object features and labels, thereby disentangling spurious correlations. 2) From the feature perspective, the simple Tail Interaction module implicitly enhances potential correlations among all samples from all source domains, facilitating the acquisition of domain-invariant representations across multiple domains for the model. The experimental results show that Convolutional Neural Networks (CNNs) or Multi-Layer Perceptrons (MLPs) with a strong baseline embedded with these two modules can achieve superior results, e.g., an average accuracy of 92.30% on Digits-DG.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (82)
  1. Invariant risk minimization. arXiv preprint arXiv, 1907.02893, 2019.
  2. Improving vision transformers by revisiting high-frequency components. arXiv preprint arXiv, 2204.00993, 2022.
  3. Recognition in terra incognita. In ECCV, pages 456–473, 2018.
  4. Rubi: Reducing unimodal biases for visual question answering. In NeurIPS, pages 839–850, 2019.
  5. Frequency domain image translation: More photo-realistic, Better identity-preserving. pages 13910–13920, 2020.
  6. Domain generalization by solving jigsaw puzzles. In CVPR, pages 2224–2233, 2019.
  7. SWAD: Domain generalization by seeking flat minima. In NeurIPS, 2021.
  8. Domain generalization by mutual-information regularization with pre-trained models. In ECCV, pages 440–457, 2022.
  9. Learning to balance specificity and invariance for in and out of domain generalization. In ECCV, pages 301–318, 2020.
  10. SMOTE: synthetic minority over-sampling technique. J. Artif. Intell. Res., 16:321–357, 2002.
  11. Compound domain generalization via meta-knowledge encoding. In CVPR, pages 7119–7129, 2022.
  12. A style and semantic memory mechanism for domain generalization. In ICCV, pages 9164–9173, 2021.
  13. Progressive random convolutions for single domain generalization. In CVPR, pages 10312–10322, 2023.
  14. Are vision transformers robust to spurious correlations? arXiv preprint arXiv, 2203.09125, 2022.
  15. In search of lost domain generalization. In ICLR, 2021.
  16. ALOFT: A lightweight mlp-like architecture with dynamic low-frequency transform for domain generalization. In CVPR, pages 24132–24141, 2023.
  17. PCT: Point cloud transformer. Computational Visual Media, 7(2):187–199, 2021.
  18. Beyond self-attention: External attention using two linear layers for visual tasks. IEEE TPAMI, 2022.
  19. Unsupervised domain generalization by learning a bridge across domains. In CVPR, pages 5280–5290, 2022.
  20. Mask r-cnn. In ICCV, pages 2980–2988, 2017.
  21. Deep residual learning for image recognition. In CVPR, pages 770–778, 2016.
  22. BatchFormer: Learning to explore sample relationships for robust representation learning. In CVPR, pages 7256–7266, 2022.
  23. Self-challenging improves cross-domain generalization. In ECCV, pages 121–140, 2020.
  24. Style Neophile: Constantly seeking novel styles for domain generalization. In CVPR, pages 7130–7140, 2022.
  25. Learning not to learn: Training deep neural networks with biased data. In CVPR, pages 9012–9020, 2019.
  26. SelfReg: Self-supervised contrastive regularization for domain generalization. In ICCV, pages 9619–9628, 2021.
  27. Adam: A method for stochastic optimization. In ICLR, 2015.
  28. Out-of-distribution generalization via risk extrapolation (rex). In ICML, volume 139, pages 5815–5826. PMLR, 2021.
  29. Single-image depth estimation based on Fourier domain analysis. In CVPR, pages 330–339, 2018.
  30. Decompose, adjust, compose: Effective normalization by playing with frequency for domain generalization. In CVPR, pages 11776–11785, 2023.
  31. Learning to generalize: Meta-learning for domain generalization. In AAAI Conference on Artificial Intelligence, 2018.
  32. Deeper, broader and artier domain generalization. In ICCV, pages 5543–5551, 2017.
  33. Domain generalization with adversarial feature learning. In CVPR, pages 5400–5409, 2018.
  34. Domain generalization for medical imaging classification with linear-dependency regularization. In NeurIPS, pages 3118–3129, 2020.
  35. Progressive domain expansion network for single domain generalization. In CVPR, pages 224–233, 2021.
  36. A simple feature augmentation for domain generalization. In ICCV, pages 8886–8895, 2021.
  37. Deep fourier ranking quantization for semi-supervised image retrieval. IEEE Trans. Image Process., 31:5909–5922, 2022.
  38. FALCON: A Fourier transform based approach for fast and secure convolutional neural network predictions. In CVPR, pages 8702–8711, 2020.
  39. Deep domain generalization via conditional invariant adversarial networks. In ECCV, pages 624–639, 2018.
  40. Pay attention to MLPs. In NeurIPS, pages 9204–9215, 2021.
  41. A background separation method of nonuniform image segmentation. 2009 4th IEEE Conference on Industrial Electronics and Applications, pages 3049–3053, 2009.
  42. Geometric and textural augmentation for domain gap reduction. In CVPR, pages 14340–14350, 2022.
  43. Decoupled weight decay regularization. In ICLR, 2019.
  44. Domain-invariant feature exploration for domain generalization. Transactions on Machine Learning Research, 2022.
  45. Causality inspired representation learning for domain generalization. In CVPR, pages 8046–8056, 2022.
  46. Domain generalization using a mixture of multiple latent domains. In AAAI, pages 11746–11756, 2020.
  47. Attention diversification for domain generalization. In ECCV, pages 322–340, 2022.
  48. Grounding visual representations with texts for domain generalization. In ECCV, pages 37–53, 2022.
  49. Domain generalization via invariant feature representation. In ICML, pages 10–18, 2013.
  50. Learning from failure: De-biasing classifier from biased classifier. Advances in Neural Information Processing Systems, 33:20673–20684, 2020.
  51. Generalization on unseen domains via inference-time label-preserving target projections. In CVPR, pages 12924–12933, 2021.
  52. How do vision transformers work? In ICLR, 2022.
  53. Moment matching for multi-source domain adaptation. In ICCV, pages 1406–1415, 2019.
  54. Efficient domain generalization via common-specific low-rank decomposition. In ICML, pages 7728–7738, 2020.
  55. Generic reversible visible watermarking via regularized graph fourier transform coding. IEEE Trans. Image Process., 31:691–705, 2022.
  56. Correlation-aware adversarial domain adaptation and generalization. PR, 100:107124, 2020.
  57. Global filter networks for image classification. In NeurIPS, pages 980–993, 2021.
  58. Multi-adversarial discriminative deep domain generalization for face presentation attack detection. In CVPR, pages 10015–10023, 2019.
  59. Gradient matching for domain generalization. In ICLR, 2022.
  60. An investigation of critical issues in bias mitigation techniques. In IEEE/CVF Winter Conference on Applications of Computer Vision, WACV, pages 2512–2523. IEEE, 2022.
  61. Amplitude scale estimation for quantization-based watermarking. IEEE Transactions on Signal Processing, 54(11):4146–4155, 2006.
  62. Background suppression based-on wavelet transformation to detect infrared target. International Conference on Machine Learning and Cybernetics, 8:4611–4615, 2005.
  63. MLP-Mixer: An all-mlp architecture for vision. In NeurIPS, pages 24261–24272, 2021.
  64. Unbiased look at dataset bias. In CVPR, pages 1521–1528, 2011.
  65. ResMLP: Feedforward networks for image classification with data-efficient training. IEEE TPAMI, 2021.
  66. Vladimir N Vapnik. An overview of statistical learning theory. IEEE transactions on neural networks, 10(5):988–999, 1999.
  67. Generalizing to unseen domains via adversarial data augmentation. In NeurIPS, pages 5334–5344, 2018.
  68. Causal balancing for domain generalization. In ICLR, 2023.
  69. PGrad: Learning principal gradients for domain generalization. In ICLR, 2023.
  70. Collaborative optimization and aggregation for decentralized domain generalization and adaptation. In ICCV, pages 6484–6493, 2021.
  71. Generative inference network for imbalanced domain generalization. IEEE Trans. Image Process., 32:1694–1704, 2023.
  72. A Fourier-based framework for domain generalization. In CVPR, pages 14383–14392, 2021.
  73. Exploiting low-rank structure from latent domains for domain generalization. In ECCV, pages 628–643, 2014.
  74. Quasi fourier-mellin transform for affine invariant features. IEEE Trans. Image Process., 29:4114–4129, 2020.
  75. PCL: Proxy-based contrastive learning for domain generalization. In CVPR, pages 7097–7107, 2022.
  76. MVDG: A unified multi-view framework for domain generalization. In ECCV, pages 161–177, 2022.
  77. Deep stable learning for out-of-distribution generalization. In CVPR, pages 5372–5382, 2021.
  78. Exact feature distribution matching for arbitrary style transfer and domain generalization. In CVPR, pages 8035–8045, 2022.
  79. FAMLP: A frequency-aware mlp-like architecture for domain generalization. arXiv preprint arXiv, 2203.12893, 2022.
  80. Learning to generate novel domains for domain generalization. In ECCV, pages 561–578, 2020.
  81. Deep domain-adversarial image generation for domain generalisation. In AAAI, pages 13025–13032, 2020.
  82. Attention consistency on visual corruptions for single-source domain generalization. In CVPRW, pages 4164–4173, 2022.
User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (5)
  1. Na Wang (147 papers)
  2. Lei Qi (84 papers)
  3. Jintao Guo (9 papers)
  4. Yinghuan Shi (79 papers)
  5. Yang Gao (761 papers)
Citations (2)
View on Semantic Scholar

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now