pith. sign in

arxiv: 2606.25784 · v1 · pith:IKAIZKHDnew · submitted 2026-06-24 · 💻 cs.CV

S²-FracMix: Label-Preserving Self-Saliency Mixup Augmentation

Khawar Islam , Arif Mahmood , Xin Jin , Naveed Akhtar This is my paper

Pith reviewed 2026-06-25 20:53 UTC · model grok-4.3

classification 💻 cs.CV
keywords data augmentationmixupself-saliencyfractal patternsimage classificationmodel robustnessscale invariancetransfer learning
0
0 comments X

The pith

Mixing multi-scale salient patches inside one image plus fractal self-similarity yields label-consistent augmentations that reach state-of-the-art across seven vision benchmarks.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper sets out to show that data augmentation can be performed entirely within a single image by pulling out salient patches at several scales and pasting them into non-salient areas of the same picture. Adding a second step that overlays self-similar fractal patterns inside the salient zones lets the model learn both fractal and ordinary structures from the identical sample. A reader would care because conventional mixup methods blend different images and often destroy the original meaning, while this intra-image route avoids that cost and the associated label noise. The authors back the approach with both a theoretical argument and results on classification, robustness, calibration, detection, and transfer tasks.

Core claim

S²-FracMix enables simultaneous learning from fractal and non-fractal structures within a single image, yielding a targeted and structurally coherent augmentation strategy that achieves state-of-the-art performance in extensive evaluation with seven benchmarks across classification, robustness, calibration, object detection, and transfer learning tasks.

What carries the argument

S²-FracMix, the combination of self-saliency patch extraction and reinsertion with adaptive-ratio fractal pattern injection into salient regions of the same image.

If this is right

  • Models learn scale-invariant features without semantic disruption from cross-sample mixing.
  • The method reduces computational overhead compared with methods that interpolate across multiple images.
  • Performance improves on both coarse and fine-grained classification tasks.
  • Gains appear in robustness, calibration, object detection, and transfer learning settings.
  • A single unified framework handles both fractal and non-fractal content inside each training example.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The intra-image constraint may encourage networks to discover features that are more stable under changes in object scale or viewpoint.
  • Similar within-sample mixing could be tested on modalities such as video or point clouds where cross-sample blending also risks semantic damage.
  • The explicit use of self-similarity patterns suggests a route for injecting natural-image priors directly into augmentation pipelines.

Load-bearing premise

Extracting multi-scale salient patches and reinserting them into non-salient regions of the same image produces samples that stay label-consistent yet are hard enough to drive scale-invariant feature learning.

What would settle it

If models trained with S²-FracMix show no accuracy, robustness, or calibration gains over standard mixup baselines when evaluated on the same seven benchmarks, the claimed advantage would be refuted.

Figures

Figures reproduced from arXiv: 2606.25784 by Arif Mahmood, Khawar Islam, Naveed Akhtar, Xin Jin.

Figure 1
Figure 1. Figure 1: (Left) Performance of ResNet-18 trained with various augmentation methods for 200 epochs on CIFAR100. (Right) Representative augmentation samples created by different methods. The samples get constructed with source and target images. methods improve robustness to unseen data, mitigate model collapse [30,58,61], and enhance resilience to distribution shifts [29, 45, 48]. A central objective of modern augme… view at source ↗
Figure 2
Figure 2. Figure 2: Comparisons of total training time vs. top-1 accuracy of ResNet-18 on CIFAR-100 dataset with RTX 3090. Within the broader task-agnostic mixup paradigm, samples are typically generated using random pairs of training instances [65]. Representative methods such as CutMix [64], Manifold Mixup [55], AlignMixup [54], and ResizeMix [51] adopt different mixing strategies to con￾struct synthetic training examples. … view at source ↗
Figure 3
Figure 3. Figure 3: Augmentation samples of S 2 - FracMix are intricate composition of frac￾tal and non-fractal patterns interpolated at multiple scales, all while retaining original image semantics. Let D = {(Ii , yi)} N i=1 represent the training dataset, where Ii ∈ R c×h×w is an input image with c channels, height h, and width w, and yi is its corresponding one-hot encoded label. Self Saliency mixup generates an aug￾mented… view at source ↗
Figure 4
Figure 4. Figure 4: Calibration plots of S 2 -FracMix on CIFAR-100 using ResNet18. 84.42%, 89.84%, outperforming the AdAutoMix by 1.06%, 1.08%. On the Stanford-Cars, it achieves 90.85%, 92.86%, exceeding the baseline by 1.20%, 1.48%. These results demonstrate that S 2 -FracMix improves fine-tuning per￾formance over baseline and recent SOTA methods across different datasets. 5.4 Robustness [PITH_FULL_IMAGE:figures/full_fig_p0… view at source ↗
Figure 5
Figure 5. Figure 5: Ablation of hyperparameters t threshold and λ for fractal mixing of S 2 - FracMix on CIFAR100 [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: (Top) Grad-CAM [52] visualization on augmented images and (bottom) t-SNE visualization of ResNet18 trained from scratch. Feature Representation. Finally, we compare the trained models by visualiz￾ing the feature representation of S 2 -FracMix and SOTA methods in [PITH_FULL_IMAGE:figures/full_fig_p015_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Top-1 accuracy on Stanford-Cars when varying the fractal library size from 100 to 500 images. The curve is essentially flat between 100 and 200 fractals (90.54%), rises modestly through 300 (90.55%), and plateaus from 400 onward at 90.56%. The saturation occurs well before 500 fractals, which suggests that the role of the library is to provide enough self-similar variation to act as structured noise rather… view at source ↗
read the original abstract

Data augmentation is known to improve generalization of deep visual models. Recent methods favor mixup strategies that generate interpolated samples to improve model performance. However, these techniques not only incur significant computational overhead, they also lead to semantic disruption of augmentation data due to cross-sample mixing. We first propose Self-Saliency ($S^2$) Mixup, which constructs challenging yet label-consistent samples by extracting multi-scale salient patches and reinserting them into non-salient regions of the same image. This promotes scale-invariant feature learning while avoiding cross-sample interference. To further enhance model robustness, we introduce FracMix, a mixing scheme that injects self-similarity patterns into salient regions using adaptive ratios. Collectively, our unified framework, $S^{2}$-FracMix, enables simultaneous learning from fractal and non-fractal structures within a single image, yielding a targeted and structurally coherent augmentation strategy. We theoretically analyze the advantage of our technique, and empirically establish its superiority over the existing methods by achieving state-of-the-art performance in extensive evaluation with seven benchmarks across classification (coarse and fine-grained), robustness, calibration, object detection, and transfer learning tasks. Project page is available at \href{https://fracmix-data-augmentation.github.io/}{fracmix-data-augmentation.github.io}

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 3 minor

Summary. The paper proposes S²-FracMix, a data augmentation framework for visual models. It introduces Self-Saliency (S²) Mixup, which extracts multi-scale salient patches from a single image and reinserts them into non-salient regions to create challenging yet label-preserving samples that encourage scale-invariant features without cross-sample mixing. It further adds FracMix, which adaptively injects fractal self-similarity patterns into salient regions. The unified method is claimed to support simultaneous learning from fractal and non-fractal structures, backed by a theoretical analysis of its advantages, and shown to achieve state-of-the-art results across seven benchmarks spanning classification (coarse/fine-grained), robustness, calibration, object detection, and transfer learning.

Significance. If the empirical results and theoretical analysis hold, the method offers a computationally lighter and semantically safer alternative to cross-sample mixup techniques by operating intra-image while incorporating fractal structures for improved robustness and scale invariance. This could meaningfully advance augmentation strategies in computer vision, particularly where label consistency and structural coherence are priorities.

minor comments (3)
  1. The abstract states that a theoretical analysis is provided, but the manuscript should explicitly state the key assumptions (e.g., on patch saliency extraction or fractal ratio adaptation) in §3 or §4 to allow readers to assess the scope of the claimed advantage.
  2. In the experimental section, the paper should report standard deviations or error bars for the SOTA claims across the seven benchmarks, as single-run results can be sensitive to random seeds in augmentation methods.
  3. Figure captions and method diagrams would benefit from clearer annotation of the adaptive ratio computation in FracMix and the multi-scale patch extraction process in S² Mixup to improve reproducibility.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive summary, significance assessment, and recommendation of minor revision. The report correctly captures the core contributions of S²-FracMix, including the intra-image self-saliency mixing and fractal pattern injection for label-preserving augmentation. No major comments were raised in the report.

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper proposes a constructive data-augmentation pipeline (Self-Saliency Mixup followed by adaptive fractal injection) whose central claims rest on the explicit algorithmic construction and subsequent empirical evaluation across benchmarks. No equations, first-principles derivations, or fitted parameters are presented that reduce by definition to the method's own inputs; the mentioned theoretical analysis is not shown to rely on self-citation chains or uniqueness theorems imported from the authors' prior work. The approach is therefore self-contained against external benchmarks and receives the default non-circularity finding.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Based solely on the abstract, no free parameters, axioms, or invented entities are explicitly stated or derivable; the method description implies unstated choices for saliency extraction and mixing ratios but provides no details.

pith-pipeline@v0.9.1-grok · 5770 in / 1358 out tokens · 62029 ms · 2026-06-25T20:53:26.908826+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

67 extracted references · 13 canonical work pages · 4 internal anchors

  1. [1]

    Data Augmentation Generative Adversarial Networks

    Antoniou, A., Storkey, A., Edwards, H.: Data augmentation generative adversarial networks. arXiv preprint arXiv:1711.04340 (2017)

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  2. [2]

    Pattern Recognition109, 107594 (2021)

    Baek, K., Bang, D., Shim, H.: Gridmix: Strong regularization through local context mapping. Pattern Recognition109, 107594 (2021)

    2021

  3. [3]

    In: European Conference on Computer Vision

    Bai, J., Yuan, L., Xia, S.T., Yan, S., Li, Z., Liu, W.: Improving vision transformers by revisiting high-frequency components. In: European Conference on Computer Vision. pp. 1–18. Springer (2022)

    2022

  4. [4]

    ACM Computing Surveys57(2), 1–38 (2024)

    Cao, C., Zhou, F., Dai, Y., Wang, J., Zhang, K.: A survey of mix-based data aug- mentation: Taxonomy, methods, applications, and explainability. ACM Computing Surveys57(2), 1–38 (2024)

    2024

  5. [5]

    Journal of Machine Learning Research23(325), 1–31 (2022)

    Carratino, L., Cissé, M., Jenatton, R., Vert, J.P.: On mixup regularization. Journal of Machine Learning Research23(325), 1–31 (2022)

    2022

  6. [6]

    Ad- vances in neural information processing systems13(2000)

    Chapelle, O., Weston, J., Bottou, L., Vapnik, V.: Vicinal risk minimization. Ad- vances in neural information processing systems13(2000)

    2000

  7. [7]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Chen, J.N., Sun, S., He, J., Torr, P.H., Yuille, A., Bai, S.: Transmix: Attend to mix for vision transformers. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 12135–12144 (2022)

    2022

  8. [8]

    Improved Baselines with Momentum Contrastive Learning

    Chen, X., Fan, H., Girshick, R., He, K.: Improved baselines with momentum con- trastive learning. arXiv preprint arXiv:2003.04297 (2020)

    work page internal anchor Pith review Pith/arXiv arXiv 2003

  9. [9]

    Chen,X.,He,K.:Exploringsimplesiameserepresentationlearning.In:Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 15750–15758 (2021)

    2021

  10. [10]

    Sensors21(24), 8444 (2021)

    Choi, J., Lee, C., Lee, D., Jung, H.: Salfmix: a novel single image-based data augmentation technique using a saliency map. Sensors21(24), 8444 (2021)

    2021

  11. [11]

    A Downsampled Variant of ImageNet as an Alternative to the CIFAR datasets

    Chrabaszcz, P., Loshchilov, I., Hutter, F.: A downsampled variant of imagenet as an alternative to the cifar datasets. arXiv preprint arXiv:1707.08819 (2017)

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  12. [12]

    Deng, J., Dong, W., Socher, R., Li, L., Li, K., Li, F.: Imagenet: A large-scale hierarchical image database (2009)

    2009

  13. [13]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Ghiasi, G., Cui, Y., Srinivas, A., Qian, R., Lin, T.Y., Cubuk, E.D., Le, Q.V., Zoph, B.: Simple copy-paste is a strong data augmentation method for instance segmentation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 2918–2928 (2021)

    2021

  14. [14]

    In:ProceedingsoftheAAAIConferenceonArtificialIntelligence.vol.33,pp.3714– 3722 (2019)

    Guo,H.,Mao,Y.,Zhang,R.:Mixupaslocallylinearout-of-manifoldregularization. In:ProceedingsoftheAAAIConferenceonArtificialIntelligence.vol.33,pp.3714– 3722 (2019)

    2019

  15. [15]

    arXiv preprint arXiv:2205.15955 (2022)

    Han, J., Petersson, L., Li, H., Reid, I.: Cropmix: Sampling a rich input distribution via multi-scale cropping. arXiv preprint arXiv:2205.15955 (2022)

    work page arXiv 2022

  16. [16]

    arXiv preprint arXiv:2002.12047 (2020)

    Harris, E., Marcu, A., Painter, M., Niranjan, M., Prügel-Bennett, A., Hare, J.: Fmix: Enhancing mixed sample data augmentation. arXiv preprint arXiv:2002.12047 (2020)

    work page arXiv 2002

  17. [17]

    He,K.,Zhang,X.,Ren,S.,Sun,J.:Deepresiduallearningforimagerecognition.In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 770–778 (2016)

    2016

  18. [18]

    Proceedings of the International Conference on Learning Representations (2019)

    Hendrycks,D.,Dietterich,T.:Benchmarkingneuralnetworkrobustnesstocommon corruptions and perturbations. Proceedings of the International Conference on Learning Representations (2019)

    2019

  19. [19]

    In: International Conference on Learning Representations (ICLR) (2020) S2-FracMix 17

    Hendrycks, D., Mu, N., Cubuk, E.D., Zoph, B., Gilmer, J., Lakshminarayanan, B.: Augmix: A simple data processing method to improve robustness and uncertainty. In: International Conference on Learning Representations (ICLR) (2020) S2-FracMix 17

    2020

  20. [20]

    In: Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Hendrycks, D., Zou, A., Mazeika, M., Tang, L., Li, B., Song, D., Steinhardt, J.: Pixmix: Dreamlike pictures comprehensively improve safety measures. In: Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 16783–16792 (2022)

    2022

  21. [21]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Hong, M., Choi, J., Kim, G.: Stylemix: Separating content and style for enhanced data augmentation. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 14862–14870 (2021)

    2021

  22. [22]

    In: 2007 IEEE Conference on computer vision and pattern recognition

    Hou, X., Zhang, L.: Saliency detection: A spectral residual approach. In: 2007 IEEE Conference on computer vision and pattern recognition. pp. 1–8. Ieee (2007)

    2007

  23. [23]

    Huang, S., Wang, X., Tao, D.: Snapmix: Semantically proportional mixing for augmentingfine-graineddata.In:ProceedingsoftheAAAIConferenceonArtificial Intelligence (2021)

    2021

  24. [24]

    Advances in Neural Information Processing Systems36, 63660–63673 (2023)

    Huang, Z., Bao, X., Zhang, N., Zhang, Q., Tu, X., Wu, B., Yang, X.: Ipmix: Label- preserving data augmentation method for training robust classifiers. Advances in Neural Information Processing Systems36, 63660–63673 (2023)

    2023

  25. [25]

    In: ICLR 2025 Workshop on Navigating and Addressing Data Prob- lems for Foundation Models

    Islam, K., AKHTAR, N.: Context-guided responsible data augmentation with dif- fusion models. In: ICLR 2025 Workshop on Navigating and Addressing Data Prob- lems for Foundation Models

    2025

  26. [26]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Islam, K., Zaheer, M.Z., Mahmood, A., Nandakumar, K.: Diffusemix: Label- preserving data augmentation with diffusion models. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 27621– 27630 (2024)

    2024

  27. [27]

    arXiv preprint arXiv:2412.02366 (2024)

    Islam, K., Zaheer, M.Z., Mahmood, A., Nandakumar, K., Akhtar, N.: Genmix: Effective data augmentation with generative diffusion model image editing. arXiv preprint arXiv:2412.02366 (2024)

    work page arXiv 2024

  28. [28]

    arXiv preprint arXiv:2510.23479 (2025)

    Jin, X., Li, S., Jian, S., Yu, K., Wang, H.: Mergemix: A unified augmen- tation paradigm for visual and multi-modal understanding. arXiv preprint arXiv:2510.23479 (2025)

    work page arXiv 2025

  29. [29]

    arXiv preprint arXiv:2409.05202 (2024)

    Jin, X., Zhu, H., Li, S., Wang, Z., Liu, Z., Yu, C., Qin, H., Li, S.Z.: A survey on mixup augmentations and beyond. arXiv preprint arXiv:2409.05202 (2024)

    work page arXiv 2024

  30. [30]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Kang, M., Kim, S.: Guidedmixup: an efficient mixup strategy guided by saliency maps. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 37, pp. 1096–1104 (2023)

    2023

  31. [31]

    In: International Conference on Machine Learning

    Kim, J.H., Choo, W., Song, H.O.: Puzzle mix: Exploiting saliency and local statis- tics for optimal mixup. In: International Conference on Machine Learning. pp. 5275–5285. PMLR (2020)

    2020

  32. [32]

    In: Advances in Neural Information Processing Systems (NeurIPS) (2021)

    Kim, J.H., Park, J., Hwang, S.: Co-mixup: Saliency-guided joint mixup with su- permodular diversity. In: Advances in Neural Information Processing Systems (NeurIPS) (2021)

    2021

  33. [33]

    In: 4th International IEEE Workshop on 3D Representation and Recognition (3dRR-13)

    Krause, J., Stark, M., Deng, J., Fei-Fei, L.: 3d object representations for fine- grained categorization. In: 4th International IEEE Workshop on 3D Representation and Recognition (3dRR-13). Sydney, Australia (2013)

    2013

  34. [34]

    Master’s thesis, University of Tront (2009)

    Krizhevsky, A.: Learning multiple layers of features from tiny images. Master’s thesis, University of Tront (2009)

    2009

  35. [35]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops

    Lee, J.H., Zaheer, M.Z., Astrid, M., Lee, S.I.: Smoothmix: a simple yet effective data augmentation to train robust classifiers. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops. pp. 756–757 (2020)

    2020

  36. [36]

    arXiv preprint arXiv:2111.15454 (2021) 18 K

    Li, S., Liu, Z., Wang, Z., Wu, D., Liu, Z., Li, S.Z.: Boosting discrimina- tive visual representation learning with scenario-agnostic mixup. arXiv preprint arXiv:2111.15454 (2021) 18 K. Islam et al

    work page arXiv 2021

  37. [37]

    Li, S., Wang, Z., Liu, Z., Wu, D., Tan, C., Jin, W., Li, S.Z.: Openmixup: A com- prehensive mixup benchmark for visual classification (2022)

    2022

  38. [38]

    Multimedia Tools and Applications84(8), 4343–4359 (2025)

    Liu, X., Shen, F., Zhao, J., Nie, C.: Randomix: A mixed sample data augmentation method with multiple mixed modes. Multimedia Tools and Applications84(8), 4343–4359 (2025)

    2025

  39. [39]

    Liu, Z., Lin, Y., Cao, Y., Hu, H., Wei, Y., Zhang, Z., Lin, S., Guo, B.: Swin transformer:Hierarchicalvisiontransformerusingshiftedwindows.In:Proceedings of the IEEE/CVF international conference on computer vision. pp. 10012–10022 (2021)

    2021

  40. [40]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Liu, Z., Mao, H., Wu, C.Y., Feichtenhofer, C., Darrell, T., Xie, S.: A convnet for the 2020s. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 11976–11986 (2022)

    2022

  41. [41]

    Advances in Neural Information Processing Systems 36(2024)

    Liu, Z., Li, S., Wang, G., Wu, L., Tan, C., Li, S.Z.: Harnessing hard mixed samples with decoupled regularizer. Advances in Neural Information Processing Systems 36(2024)

    2024

  42. [42]

    In: European Conference on Computer Vision

    Liu, Z., Li, S., Wu, D., Liu, Z., Chen, Z., Wu, L., Li, S.Z.: Automix: Unveiling the power of mixup for stronger classifiers. In: European Conference on Computer Vision. pp. 441–458. Springer (2022)

    2022

  43. [43]

    Journal of machine learning research9(11) (2008)

    Van der Maaten, L., Hinton, G.: Visualizing data using t-sne. Journal of machine learning research9(11) (2008)

    2008

  44. [44]

    Fine-Grained Visual Classification of Aircraft

    Maji, S., Rahtu, E., Kannala, J., Blaschko, M., Vedaldi, A.: Fine-grained visual classification of aircraft. arXiv preprint arXiv:1306.5151 (2013)

    work page internal anchor Pith review Pith/arXiv arXiv 2013

  45. [45]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Parast, A.Y., Azam, B., Akhtar, N.: Ddb: Diffusion driven balancing to address spurious correlations. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 17526–17535 (2025)

    2025

  46. [46]

    arXiv preprint arXiv:2509.25178 (2025)

    Parast, A.Y., Hosseini, P., Asadollahzadeh, H., Moakhar, A.S., Azam, B., Feizi, S., Akhtar, N.: Ghost: Hallucination-inducing image generation for multimodal llms. arXiv preprint arXiv:2509.25178 (2025)

    work page arXiv 2025

  47. [47]

    arXiv preprint arXiv:2603.29313 (2026)

    Parast, A.Y., Islam, K., Won, S., Azam, B., Akhtar, N.: Hsfm: Hard-set-guided feature-space meta-learning for robust classification under spurious correlations. arXiv preprint arXiv:2603.29313 (2026)

    work page arXiv 2026

  48. [48]

    Advances in Neural Information Processing Systems35, 14608–14622 (2022)

    Pinto, F., Yang, H., Lim, S.N., Torr, P., Dokania, P.: Using mixup as a regularizer can surprisingly improve accuracy & out-of-distribution robustness. Advances in Neural Information Processing Systems35, 14608–14622 (2022)

    2022

  49. [49]

    Qin,H.,Jin,X.,Jiang,Y.,El-Yacoubi,M.,Gao,X.:Adversarialautomixup.In:The Twelfth International Conference on Learning Representations. pp. OpenReview, Spotlight (2024)

    2024

  50. [50]

    In: European Conference on Computer Vision

    Qin, H., Jin, X., Zhu, H., Liao, H., El-Yacoubi, M.A., Gao, X.: Sumix: Mixup with semantic and uncertain information. In: European Conference on Computer Vision. pp. 70–88. Springer (2025)

    2025

  51. [51]

    arXiv preprint arXiv:2012.11101 (2020)

    Qin, J., Fang, J., Zhang, Q., Liu, W., Wang, X., Wang, X.: Resizemix: Mixing data with preserved object information and true labels. arXiv preprint arXiv:2012.11101 (2020)

    work page arXiv 2012

  52. [52]

    In- ternational journal of computer vision128, 336–359 (2020)

    Selvaraju, R.R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., Batra, D.: Grad- cam: visual explanations from deep networks via gradient-based localization. In- ternational journal of computer vision128, 336–359 (2020)

    2020

  53. [53]

    In: International Con- ference on Learning Representations (2020) S2-FracMix 19

    Uddin, A.S., Monira, M.S., Shin, W., Chung, T., Bae, S.H.: Saliencymix: A saliency guided data augmentation strategy for better regularization. In: International Con- ference on Learning Representations (2020) S2-FracMix 19

    2020

  54. [54]

    Venkataramanan, S., Kijak, E., Amsaleg, L., Avrithis, Y.: Alignmixup: Improving representationsbyinterpolatingalignedfeatures.In:ProceedingsoftheIEEE/CVF conference on computer vision and pattern recognition. pp. 19174–19183 (2022)

    2022

  55. [55]

    In: International Conference on Machine Learning

    Verma, V., Lamb, A., Beckham, C., Najafi, A., Mitliagkas, I., Lopez-Paz, D., Ben- gio, Y.: Manifold mixup: Better representations by interpolating hidden states. In: International Conference on Machine Learning. pp. 6438–6447. PMLR (2019)

    2019

  56. [56]

    Wah, C., Branson, S., Welinder, P., Perona, P., Belongie, S.: The Caltech-UCSD Birds-200-2011 Dataset. Tech. Rep. CNS-TR-2011-001, California Institute of Technology (2011)

    2011

  57. [57]

    In: ICASSP 2020-2020 IEEEInternational Conference on Acoustics, Speech and Signal Processing (ICASSP)

    Walawalkar, D., Shen, Z., Liu, Z., Savvides, M.: Attentive cutmix: An enhanced data augmentation approach for deep learning based image classification. In: ICASSP 2020-2020 IEEEInternational Conference on Acoustics, Speech and Signal Processing (ICASSP). pp. 3642–3646. IEEE (2020)

    2020

  58. [58]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Wang, Z., Wei, L., Wang, T., Chen, H., Hao, Y., Wang, X., He, X., Tian, Q.: Enhance image classification via inter-class image mixup with diffusion model. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 17223–17233 (2024)

    2024

  59. [59]

    Neural Networks190, 107611 (2025)

    Won, S., Bae, S.H., Kim, S.T.: Effects of mixed sample data augmentation on interpretability of neural networks. Neural Networks190, 107611 (2025)

    2025

  60. [60]

    Image and Vision Com- puting p

    Won, S., Kim, H.B., Ahn, Y.H., Lee, H.J., Kim, S.T.: Understanding adversarial robustness of deep neural networks via decision reliance. Image and Vision Com- puting p. 105743 (2025)

    2025

  61. [61]

    In: Proceedings of the IEEE/CVF international conference on com- puter vision

    Xiao, H., Zheng, W., Zhu, Z., Zhou, J., Lu, J.: Token-label alignment for vision transformers. In: Proceedings of the IEEE/CVF international conference on com- puter vision. pp. 5495–5504 (2023)

    2023

  62. [62]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Xie,S.,Girshick,R.,Dollár,P.,Tu,Z.,He,K.:Aggregatedresidualtransformations for deep neural networks. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 1492–1500 (2017)

    2017

  63. [63]

    Advances in Neural Information Processing Systems35, 8427–8440 (2022)

    Yang, L., Li, X., Zhao, B., Song, R., Yang, J.: Recursivemix: Mixed learning with history. Advances in Neural Information Processing Systems35, 8427–8440 (2022)

    2022

  64. [64]

    In: International Con- ference on Computer Vision (ICCV) (2019)

    Yun, S., Han, D., Oh, S.J., Chun, S., Choe, J., Yoo, Y.: Cutmix: Regularization strategy to train strong classifiers with localizable features. In: International Con- ference on Computer Vision (ICCV) (2019)

    2019

  65. [65]

    In: International Conference on Learning Representations (ICLR) (2018)

    Zhang, H., Cisse, M., Dauphin, Y.N., Lopez-Paz, D.: mixup: Beyond empirical risk minimization. In: International Conference on Learning Representations (ICLR) (2018)

    2018

  66. [66]

    Advances in Neural Infor- mation Processing Systems33, 14435–14447 (2020)

    Zhao, L., Liu, T., Peng, X., Metaxas, D.: Maximum-entropy adversarial data aug- mentation for improved generalization and robustness. Advances in Neural Infor- mation Processing Systems33, 14435–14447 (2020)

    2020

  67. [67]

    In: Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XXVII 16

    Zoph, B., Cubuk, E.D., Ghiasi, G., Lin, T.Y., Shlens, J., Le, Q.V.: Learning data augmentation strategies for object detection. In: Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XXVII 16. pp. 566–583. Springer (2020) 20 K. Islam et al. Overview of the Appendices This supplementary material compleme...

    2020