Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

Fine-Grained Image Recognition from Scratch with Teacher-Guided Data Augmentation

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read High-performance fine-grained image recognition can be trained entirely from scratch, the paper argues, using teacher-guided augmentation that distills part-attention knowledge into a randomly initialized student.

desk verdict Useful distillation recipe and backbone designs, but the 'from scratch' claim does not survive contact with the pretrained teacher. read the letter →

arxiv 2507.12157 v1 pith:2MWILEDK submitted 2025-07-16 cs.CV

Edwin Arkel Rios , Fernando Mikael , Oswin Gosal , Femiloye Oyerinde , Hao-Chun Liang , Bo-Cheng Lai , Min-Chun Hu This is my paper
classification cs.CV
keywords fine-grainedimagerecognitiontrainingfromscratchknowledgedistillationdataaugmentationpartattentionmapslow-resolutionvisiontransformersefficientinference
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper sets out to show that fine-grained image recognition, which must tell apart near-identical bird species, car models, or aircraft, does not require backbones pretrained on ImageNet. It proposes Teacher-Guided Data Augmentation (TGDA), a two-stage process in which a teacher trained with part-attention maps first generates augmented crops of each image, and the frozen teacher's soft predictions then supervise a student trained from random initialization. The claim is that TGDA-trained students match or beat pretrained state-of-the-art methods on three benchmarks, and that this unlocks backbones designed specifically for the task: LRNets for low-resolution input and ViTFS, a batch-normalized vision transformer family. If true, practitioners could drop pretraining entirely, saving data and compute and enabling deployment-friendly architectures.

What carries the argument

The load-bearing mechanism is the Part Attention Module (PAM), a learnable convolutional unit that produces multiple attention maps highlighting semantically meaningful object regions. These maps are used in two data-aware augmentations: attention cropping, which zooms into discriminative regions, and attention dropping, which masks them to force complementary feature learning. The same maps, combined with the teacher's soft predictions on original and augmented views, give the student rich supervision without any pretrained weight initialization. The framework is two-stage: the teacher is trained first with its PAM, then frozen while supervising the randomly initialized student; prior one-stage co-training, in which an untrained student also teaches the teacher, is shown to be harmful.

What would settle it

Train the TGDA pipeline with a teacher that is itself trained from random initialization on the target dataset only, and compare student accuracy; if it falls to the low single-digit levels of the no-teacher ablation, the reported gains come from the teacher's ImageNet-pretrained knowledge rather than from the mechanism itself.

Watch

Extended reading notes

Core claim

The paper's central claim is that high-performance fine-grained image recognition can be trained entirely from scratch, provided the student is guided by a teacher that has learned where discriminative parts are. In TGDA, a fine-grained-aware teacher built on a part attention module produces attention maps; these maps drive attention cropping and attention dropping to create diverse augmented views, and the teacher's soft labels on both original and augmented images form the distillation signal for a randomly initialized student. On ResNet-18/34 at 448x448, TGDA outperforms the previous distillation-based state of the art by up to 2.4 accuracy points; with the dedicated LRNet backbone at 128x128 it beats prior low-resolution methods by up to 23 points while using up to 20.6x fewer parameters. The paper also reports that ViTFS-T, trained with TGDA on only a few thousand images, matches a ViT-B pretrained on ImageNet-21k with 15.3x fewer trainable parameters.

Load-bearing premise

The claim that TGDA removes reliance on pretraining assumes that the pretrained knowledge entering through the teacher, which is fine-tuned from an ImageNet-pretrained ResNet-101, does not count as reliance on pretraining.

Editorial extensions

If this is right

  • Fine-grained recognition systems can be built without ImageNet pretraining, so backbones can be chosen for deployment constraints such as resolution, parameters, and hardware support rather than for what pretrained weights happen to exist.
  • Small task-specific backbones like LRNets become practical for low-resolution input, where ImageNet-pretrained architectures lose the fine detail needed to separate near-identical classes.
  • Batch-normalized vision transformers like ViTFS can be trained from random initialization, making transformer inference faster on hardware that lacks optimized LayerNorm support.
  • The reported gains come with far less training data and compute, lowering the barrier to entry for resource-constrained labs and applications.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Inference: because the teacher is itself ImageNet-pretrained, the honest reading of "from scratch" is narrow: TGDA replaces direct pretraining of the deployable student with transfer of pretrained knowledge through the teacher, and the paper does not demonstrate that a from-scratch teacher would work as well.
  • Inference: the method's dependence on part attention suggests it should transfer best to domains where classes differ in localized parts; for categories that differ mainly in global appearance, attention-based cropping and dropping may add less.
  • Inference: a natural extension is to test TGDA on medical or agricultural fine-grained tasks, where ImageNet pretraining is often mismatched to the imaging modality and a task-designed backbone could show larger gains.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes TGDA, a two-stage knowledge-distillation framework for fine-grained image recognition (FGIR) that aims to train student models from random initialization using a fine-grained-aware teacher to generate attention-based augmentations and soft supervision. The authors introduce two task-specific backbone families, LRNets for low-resolution FGIR and ViTFS for hardware-efficient vision transformers, and report that these from-scratch students outperform pretrained baselines on FGVC-Aircraft, Stanford-Cars, and CUB-200-2011 in both high- and low-resolution settings.

Significance. If the claims were fully supported, the work would be significant: it would challenge the field's reliance on ImageNet pretraining for FGIR and open the door to task-specific architectures. The paper has several strengths: a clear two-stage framework, a thoughtful reassessment of backbone design for low-resolution inputs, hardware-aware transformer modifications, and extensive comparisons across three benchmarks. However, the central 'from scratch' claim is undercut by the experimental design, as detailed in the major comments, and the present version does not establish the claimed independence from pretraining.

major comments (3)
  1. [§1, §2.2, §4] The central claim that TGDA enables training 'from scratch' without reliance on pretrained weights is contradicted by the method itself. Section 4 states that 'ResNet-101 teachers are fine-tuned for 50 epochs with SGD based on the CAL recipe'; fine-tuning conventionally starts from an ImageNet-pretrained checkpoint, and the paper never states otherwise or trains a teacher from random initialization. In Stage 2 (Section 2.2), the student's only supervision beyond ground-truth labels is the teacher's soft predictions and attention maps, so the teacher's pretrained knowledge is propagated into the from-scratch student. The claimed independence from pretraining is therefore not demonstrated; the results could be a transfer of ImageNet-trained features through the teacher. The manuscript must either train teachers from scratch and re-run all experiments, or explicitly reframe the contribution as 'from-scratch student with a pretrained teacher,' which would substantially reduce the novelty claimed in the title and abstract.
  2. [§5.3, Table 4] The ablation study does not isolate the contribution of the teacher's pretrained knowledge. The rows without 'Two-Stage' report near-random accuracies (e.g., 1.7% on Aircraft, 0.6% on Cars, 10.1% on CUB for the no-component baseline), and the jump to 89.0% on Aircraft when 'Two-Stage' is added is attributed to the two-stage training design; however, the 'Two-Stage' condition uses a fine-tuned, pretrained teacher. The table also lacks an explicit 'Teacher' column, so it is unclear whether the non-two-stage rows even include a teacher. As written, Table 4 confounds the effect of a separate teacher with the effect of pretrained initialization, and it cannot support the claim that PAM and two-stage training are sufficient for from-scratch success.
  3. [§5.1, Tables 1 and 2; §5.2, Table 3] The comparisons against pretrained baselines are not apples-to-apples. TGDA's student receives supervision from a ResNet-101 teacher that has been fine-tuned on the target dataset from ImageNet initialization, whereas the baselines are standard backbones directly fine-tuned. Thus the reported gains (up to 2.4 points over CEKD at 448×448 and up to 23 points over DSSD at 128×128) may reflect distillation from a much stronger pretrained teacher rather than a genuine advantage of from-scratch training. A control experiment in which the teacher is trained from scratch—or, at minimum, in which the student is trained with a from-scratch teacher—is needed to support the paper's conclusions. Without such a control, the headline results are consistent with the alternative hypothesis that pretraining has merely been moved into the teacher.
minor comments (5)
  1. [§4, References] Reference [6] is cited for CUB-200-2011, but [6] is an empirical study of self-supervised vision transformers; the correct CUB citation appears to be [25].
  2. [§5.2, Table 3] The parameter counts for LRNet versus standard ResNets are not reported in the table; the text refers to Figure 1, but the exact reductions should be stated to support the '20.6x less parameters' claim.
  3. [§4] The loss weight α is not reported; please provide the value and, ideally, a sensitivity analysis for β and temperature.
  4. [§5.3, Table 4] Table 4 would benefit from an explicit 'Teacher' column and from error bars or multiple seeds, given the huge variance between rows.
  5. [§5.2, Abstract] Minor wording issues: 'Tab. 3 present' should be 'Tab. 3 presents,' and 'orders of magnitude less data' should be quantified (Table 2 shows 6K images for ViTFS-T, which is roughly three orders of magnitude less than ImageNet-21k, not simply 'orders of magnitude' without specifying the base).

Circularity Check

0 steps flagged · score 1.0 of 10

No formal circularity; the 'from scratch' claim is weakened by a pretrained teacher, which is a dependency gap, not a definitional reduction.

full rationale

The derivation chain is not circular in the formal sense. The teacher (fine-tuned ResNet-101) and student (randomly initialized LRNet/ViTFS) are distinct models, and the reported top-1 accuracies are measured on held-out test images, so no fitted parameter is being renamed as a prediction and no tested quantity is constructed from itself. The closest concern is the paper's claim that TGDA enables 'training a student model from random initialization, without any reliance on external pretrained weights' (Sec. 1) alongside Section 4's statement that 'ResNet-101 teachers are fine-tuned for 50 epochs with SGD based on the CAL recipe [19]' and Section 2.2's KD loss L_total = alpha * L_CE(I_org) + beta * [L_KD(I_org) + L_KD(I_aug)], through which only the student is from scratch. This means ImageNet knowledge can enter through the teacher, so the 'from scratch' conclusion is less independent of pretraining than the prose suggests. However, this is an experimental-support and confounding issue, not a circular definition: the teacher's weights are not defined in terms of the student's held-out accuracy, and the student's result is not an identity. No load-bearing self-citation, uniqueness-import, ansatz-smuggling, or renaming pattern appears. The score reflects the minor framing overreach, not equation-level circularity.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

LRNets and ViTFS are new neural architectures, not new physical entities. No new particles, forces, mediators, or unmeasured objects are introduced. The central claim rests on a pretrained teacher and several manually chosen hyperparameters.

free parameters (3)
  • KD loss weight beta = 10
    Section 4: 'we set the loss weight hyperparameter for TGDA to beta = 10'. No sensitivity analysis is reported, so the results may depend on this manual choice.
  • distillation temperature = 7
    Section 4: 'the distillation temperature = 7'. All student training uses this value; the paper does not study sensitivity to it.
  • CE loss weight alpha = not reported
    Eq. (1) in Section 2.2 defines L_total = alpha L_CE(I_org) + beta [L_KD(I_org) + L_KD(I_aug)], but alpha is never specified, leaving the student objective incompletely defined.
assumptions (4)
  • domain assumption A fine-tuned teacher's part attention maps and soft labels provide supervision sufficient to train a student from scratch.
    Sections 2.1 and 2.2 build the entire framework on this premise. The paper never tests a teacher trained without ImageNet initialization, so this assumption is load-bearing and untested.
  • domain assumption Part attention modules from WS-DAN/CAL localize discriminative regions for fine-grained classes.
    The teacher's PAMs are used directly as augmentation drivers in Section 2.1. No analysis of failure cases or attention quality is provided.
  • domain assumption BatchNorm is better than LayerNorm for inference on target mobile hardware.
    Section 3.2 reports a single measurement on one Galaxy A53 device. The claim is hardware-specific and not broadly established.
  • domain assumption Generic regularization techniques (TrivialAugment, label smoothing, stochastic depth) transfer to from-scratch FGIR training.
    Applied in Section 4 without ablated individual contributions, so the framework's gains are not isolated from these standard tricks.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fine-Grained Image Recognition from Scratch with Teacher-Guided Data Augmentation." pith.science (2026). https://pith.science/paper/2MWILEDK

@misc{pith2026250712157,
  author       = {Pith},
  title        = {Pith review of: Fine-Grained Image Recognition from Scratch with Teacher-Guided Data Augmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2MWILEDK}},
  note         = {Machine review of arXiv:2507.12157}
}
read the original abstract

Fine-grained image recognition (FGIR) aims to distinguish visually similar sub-categories within a broader class, such as identifying bird species. While most existing FGIR methods rely on backbones pretrained on large-scale datasets like ImageNet, this dependence limits adaptability to resource-constrained environments and hinders the development of task-specific architectures tailored to the unique challenges of FGIR. In this work, we challenge the conventional reliance on pretrained models by demonstrating that high-performance FGIR systems can be trained entirely from scratch. We introduce a novel training framework, TGDA, that integrates data-aware augmentation with weak supervision via a fine-grained-aware teacher model, implemented through knowledge distillation. This framework unlocks the design of task-specific and hardware-aware architectures, including LRNets for low-resolution FGIR and ViTFS, a family of Vision Transformers optimized for efficient inference. Extensive experiments across three FGIR benchmarks over diverse settings involving low-resolution and high-resolution inputs show that our method consistently matches or surpasses state-of-the-art pretrained counterparts. In particular, in the low-resolution setting, LRNets trained with TGDA improve accuracy by up to 23\% over prior methods while requiring up to 20.6x less parameters, lower FLOPs, and significantly less training data. Similarly, ViTFS-T can match the performance of a ViT B-16 pretrained on ImageNet-21k while using 15.3x fewer trainable parameters and requiring orders of magnitudes less data. These results highlight TGDA's potential as an adaptable alternative to pretraining, paving the way for more efficient fine-grained vision systems.

Figures

Figures reproduced from arXiv: 2507.12157 by the authors.

Figure 1
Figure 1. Average vs cost trade-offs of our method compared to prior work across diverse FGIR [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Overview of the TGDA framework. TGDA is a two-stage training pipeline for fine-grained [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. How to Choose Your Teacher for Fine Grained Image Recognition

    cs.CV 2026-05 conditional novelty 6.0 of 10

    Proposes Ratio 1-2 metric for teacher selection in knowledge distillation for fine-grained image recognition, validated across 1000+ experiments showing 18% better selection and up to 17% student accuracy gains.

Reference graph

Works this paper leans on

39 extracted references · 23 canonical work pages · cited by 1 Pith paper

  1. [1]

    URL https://developer.arm.com/Processors/Ethos-U65

    Ethos-U65, . URL https://developer.arm.com/Processors/Ethos-U65

    work page

  2. [2]

    URL https://aihub.qualcomm.com

    Qualcomm AI Hub, . URL https://aihub.qualcomm.com

    work page

  3. [3]

    Jimmy Lei Ba, Jamie Ryan Kiros, and Geoffrey E. Hinton. Layer Normalization, July 2016. URL http://arxiv.org/abs/1607.06450. arXiv:1607.06450 [cs, stat]

    arXiv 2016

  4. [4]

    Better plain ViT baselines for ImageNet- 1k, May 2022

    Lucas Beyer, Xiaohua Zhai, and Alexander Kolesnikov. Better plain ViT baselines for ImageNet- 1k, May 2022. URL http://arxiv.org/abs/2205.01580. arXiv:2205.01580 [cs]

    arXiv 2022

  5. [5]

    Training Vision Transformers with only 2040 Images

    Yun-Hao Cao, Hao Yu, and Jianxin Wu. Training Vision Transformers with only 2040 Images. In Computer Vision – ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23–27, 2022, Proceedings, Part XXV , pages 220–237, Berlin, Heidelberg, October 2022. Springer-Verlag. ISBN 978-3-031-19805-2. doi: 10.1007/978-3-031-19806-9_13. URL https://doi.org/10...

    work page doi:10.1007/978-3-031-19806-9_13 2022

  6. [7]

    Vision Transformers Need Registers

    Timothée Darcet, Maxime Oquab, Julien Mairal, and Piotr Bojanowski. Vision Transformers Need Registers. October 2023. URL https://openreview.net/forum?id=2dnO3LLiJ1

    work page 2023

  7. [8]

    Look Closer to See Better: Recurrent Attention Convolutional Neural Network for Fine-Grained Image Recognition

    Jianlong Fu, Heliang Zheng, and Tao Mei. Look Closer to See Better: Recurrent Attention Convolutional Neural Network for Fine-Grained Image Recognition. In 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , pages 4476–4484, July 2017. doi: 10.1109/CVPR.2017.476

    work page doi:10.1109/cvpr.2017.476 2017

  8. [9]

    Maybank, and Dacheng Tao

    Jianping Gou, Baosheng Yu, Stephen J. Maybank, and Dacheng Tao. Knowledge Distilla- tion: A Survey. International Journal of Computer Vision , 129(6):1789–1819, June 2021. ISSN 1573-1405. doi: 10.1007/s11263-021-01453-z. URL https://doi.org/10.1007/ s11263-021-01453-z

    doi:10.1007/s11263-021-01453-z 2021

Show all 39 references
  1. [10]

    Distilling the Knowledge in a Neural Network, March 2015

    Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. Distilling the Knowledge in a Neural Network, March 2015. URL http://arxiv.org/abs/1503.02531

    2015 arXiv

  2. [11]

    Tao Hu and H. Qi. See Better Before Looking Closer: Weakly Supervised Data Aug- mentation Network for Fine-Grained Visual Classification. ArXiv, January 2019. URL https://www.semanticscholar.org/paper/See-Better-Before-Looking-Closer% 3A-Weakly-Supervised-Hu-Qi/a21d8d27caca866...

    2019

  3. [12]

    Weinberger

    Gao Huang, Yu Sun, Zhuang Liu, Daniel Sedra, and Kilian Q. Weinberger. Deep Networks with Stochastic Depth. In Bastian Leibe, Jiri Matas, Nicu Sebe, and Max Welling, editors, Computer Vision – ECCV 2016, Lecture Notes in Computer Science, pages 646–661, Cham, 2016. Springer In...

    2016 doi

  4. [13]

    SnapMix: Semantically Proportional Mixing for Augmenting Fine-grained Data

    Shaoli Huang, Xinchao Wang, and Dacheng Tao. SnapMix: Semantically Proportional Mixing for Augmenting Fine-grained Data. In Proceedings of the AAAI Conference on Artificial Intelli- gence, volume 35, pages 1628–1636, May 2021. doi: 10.1609/aaai.v35i2.16255. URL https: //ojs.aa...

    2021 doi

  5. [14]

    Batch normalization: accelerating deep network training by reducing internal covariate shift

    Sergey Ioffe and Christian Szegedy. Batch normalization: accelerating deep network training by reducing internal covariate shift. In Proceedings of the 32nd International Conference on International Conference on Machine Learning - Volume 37, ICML’15, pages 448–456, Lille, Fra...

    2015

  6. [15]

    3D Object Representations for Fine- Grained Categorization

    Jonathan Krause, Michael Stark, Jia Deng, and Li Fei-Fei. 3D Object Representations for Fine- Grained Categorization. 2013 IEEE International Conference on Computer Vision Workshops, pages 554–561, December 2013. doi: 10.1109/ICCVW.2013.77. URL http://ieeexplore. ieee.org/docu...

    2013

  7. [16]

    Dynamic semantic structure distillation for low-resolution fine-grained recognition

    Mingjiang Liang, Shaoli Huang, and Wei Liu. Dynamic semantic structure distillation for low-resolution fine-grained recognition. Pattern Recognition, 148:110216, April 2024. ISSN 0031-3203. doi: 10.1016/j.patcog.2023.110216. URL https://www.sciencedirect.com/ science/article/p...

    2024

  8. [17]

    Blaschko, and A

    Subhransu Maji, Esa Rahtu, Juho Kannala, Matthew B. Blaschko, and A. Vedaldi. Fine-Grained Visual Classification of Aircraft. ArXiv, June 2013. URL https://www.semanticscholar. org/paper/Fine-Grained-Visual-Classification-of-Aircraft-Maji-Rahtu/ 522d65a3db7431015aeaa201a7fc445...

    2013

  9. [18]

    Müller and Frank Hutter

    Samuel G. Müller and Frank Hutter. TrivialAugment: Tuning-free Yet State-of-the-Art Data Augmentation, August 2021. URL http://arxiv.org/abs/2103.10158. arXiv:2103.10158 [cs]

    2021 arXiv

  10. [19]

    Counterfactual Attention Learn- ing for Fine-Grained Visual Categorization and Re-identification

    Yongming Rao, Guangyi Chen, Jiwen Lu, and Jie Zhou. Counterfactual Attention Learn- ing for Fine-Grained Visual Categorization and Re-identification. In 2021 IEEE/CVF International Conference on Computer Vision (ICCV) , pages 1005–1014, October 2021. doi: 10.1109/ICCV48922.202...

    2021

  11. [20]

    Rethinking the Inception Architecture for Computer Vision

    Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jon Shlens, and Zbigniew Wojna. Rethinking the Inception Architecture for Computer Vision. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , pages 2818–2826, June 2016. doi: 10.1109/CVPR.2016.308. IS...

    2016 doi

  12. [21]

    Neil Thompson, Kristjan Greenewald, Keeheon Lee, and Gabriel F. Manso. The Computational Limits of Deep Learning. In Ninth Computing within Limits 2023, Virtual, June 2023. LIMITS. doi: 10.21428/bf6fb269.1f033948. URL https://limits.pubpub.org/pub/wm1lwjce

    2023 doi

  13. [22]

    Contrastive Representation Distillation

    Yonglong Tian, Dilip Krishnan, and Phillip Isola. Contrastive Representation Distillation. September 2019. URL https://openreview.net/forum?id=SkgpBJrtvS

    2019

  14. [23]

    Training data-efficient image transformers & distillation through attention

    Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, and Hervé Jégou. Training data-efficient image transformers & distillation through attention. arXiv:2012.12877 [cs], January 2021. URL http://arxiv.org/abs/2012.12877. arXiv: 2012.12877

    2012 arXiv

  15. [24]

    Building a bird recognition app and large scale dataset with citizen scientists: The fine print in fine-grained dataset collection

    Grant Van Horn, Steve Branson, Ryan Farrell, Scott Haber, Jessie Barry, Panos Ipeirotis, Pietro Perona, and Serge Belongie. Building a bird recognition app and large scale dataset with citizen scientists: The fine print in fine-grained dataset collection. 2015 IEEE Conference ...

    2015

  16. [25]

    Wah, Steve Branson, P

    C. Wah, Steve Branson, P. Welinder, P. Perona, and Serge J. Belongie. The Caltech- UCSD Birds-200-2011 Dataset. July 2011. URL https://www.semanticscholar. org/paper/The-Caltech-UCSD-Birds-200-2011-Dataset-Wah-Branson/ c069629a51f6c1c301eb20ed77bc6b586c24ce32. 9

    2011

  17. [26]

    Identification of Apple Leaf Diseases by Improved Deep Convolutional Neural Networks With an Attention Mechanism

    Peng Wang, Tong Niu, Yanru Mao, Zhao Zhang, Bin Liu, and Dongjian He. Identification of Apple Leaf Diseases by Improved Deep Convolutional Neural Networks With an Attention Mechanism. Frontiers in Plant Science, 12:723294, September 2021. ISSN 1664-462X. doi: 10.3389/fpls.2021...

    2021

  18. [27]

    Fine-Grained Image Analysis With Deep Learning: A Survey

    Xiu-Shen Wei, Yi-Zhe Song, Oisin Mac Aodha, Jianxin Wu, Yuxin Peng, Jinhui Tang, Jian Yang, and Serge Belongie. Fine-Grained Image Analysis With Deep Learning: A Survey. IEEE Transactions on Pattern Analysis and Machine Intelligence , 44(12):8927–8948, December

  19. [28]

    Early Convolutions Help Transformers See Better

    Tete Xiao, Mannat Singh, Eric Mintun, Trevor Darrell, Piotr Dollar, and Ross Gir- shick. Early Convolutions Help Transformers See Better. In Advances in Neu- ral Information Processing Systems , volume 34, pages 30392–30400. Curran Asso- ciates, Inc., 2021. URL https://proceed...

    2021

  20. [29]

    Shuo Ye, Yu Wang, Qinmu Peng, Xinge You, and C. L. Philip Chen. The Image Data and Backbone in Weakly Supervised Fine-Grained Visual Categorization: A Revisit and Further Thinking. IEEE Transactions on Circuits and Systems for Video Technology , 34(1):2–16, January 2024. ISSN ...

    2024

  21. [30]

    Data Augmentation Guided Decouple Knowledge Distillation for Low-Resolution Fine-Grained Image Classification

    Haotian Zhang, Yongliang Qiao, and Meili Wang. Data Augmentation Guided Decouple Knowledge Distillation for Low-Resolution Fine-Grained Image Classification. In Zhouchen Lin, Ming-Ming Cheng, Ran He, Kurban Ubul, Wushouer Silamu, Hongbin Zha, Jie Zhou, and Cheng-Lin Liu, edito...

  22. [31]

    CEKD:Cross ensemble knowledge distillation for augmented fine-grained data

    Ke Zhang, Jin Fan, Shaoli Huang, Yongliang Qiao, Xiaofeng Yu, and Feiwei Qin. CEKD:Cross ensemble knowledge distillation for augmented fine-grained data. Ap- plied Intelligence , 52(14):16640–16650, November 2022. ISSN 0924-669X, 1573-

    2022

  23. [32]

    Intra-class Part Swapping for Fine-Grained Image Classification

    Lianbo Zhang, Shaoli Huang, and Wei Liu. Intra-class Part Swapping for Fine-Grained Image Classification. 2021 IEEE Winter Conference on Applications of Computer Vision (WACV), pages 3208–3217, January 2021. doi: 10.1109/WACV48630.2021.00325. URL https://ieeexplore.ieee.org/do...

    2021

  24. [33]

    S3Mix: Same Category Same Semantics Mixing for Augmenting Fine-grained Images

    Zi-Chao Zhang, Zhen-Duo Chen, Zhen-Yu Xie, Xin Luo, and Xin-Shun Xu. S3Mix: Same Category Same Semantics Mixing for Augmenting Fine-grained Images. ACM Trans. Mul- timedia Comput. Commun. Appl. , 20(1):9:1–9:16, August 2023. ISSN 1551-6857. doi: 10.1145/3605892. URL https://dl...

    2023 doi

  25. [34]

    Decoupled Knowledge Distillation

    Borui Zhao, Quan Cui, Renjie Song, Yiyu Qiu, and Jiajun Liang. Decoupled Knowledge Distillation. In 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 11943–11952, New Orleans, LA, USA, June 2022. IEEE. ISBN 978-1-6654- 6946-3. doi: 10.1109/CVPR5...

    2022

  26. [35]

    Learning Multi-attention Convolutional Neural Network for Fine-Grained Image Recognition

    Heliang Zheng, Jianlong Fu, Tao Mei, and Jiebo Luo. Learning Multi-attention Convolutional Neural Network for Fine-Grained Image Recognition. In 2017 IEEE International Conference on Computer Vision (ICCV), pages 5219–5227, Venice, October 2017. IEEE. ISBN 978-1-5386- 1032-9. ...

    2017 doi

  27. [36]

    Learning Deep Features for Discriminative Localization

    Bolei Zhou, Aditya Khosla, Agata Lapedriza, Aude Oliva, and Antonio Torralba. Learning Deep Features for Discriminative Localization. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 2921–2929, June 2016. doi: 10.1109/CVPR.2016

    2016 doi

  28. [319]

    Conference Name: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) ISBN: 9781467388511 Place: Las Vegas, NV , USA Publisher: IEEE

    URL http://ieeexplore.ieee.org/document/7780688/. Conference Name: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) ISBN: 9781467388511 Place: Las Vegas, NV , USA Publisher: IEEE. 10

    2016

  29. [2022]

    doi: 10.1109/TPAMI.2021.3126648

    ISSN 0162-8828, 2160-9292, 1939-3539. doi: 10.1109/TPAMI.2021.3126648. URL https://ieeexplore.ieee.org/document/9609630/

    1939

  30. [2025]

    ISBN 978-981-97-8505-6

    Springer Nature. ISBN 978-981-97-8505-6. doi: 10.1007/978-981-97-8505-6_27

    doi

  31. [7497]

    URL https://link.springer.com/10.1007/ s10489-022-03355-0

    doi: 10.1007/s10489-022-03355-0. URL https://link.springer.com/10.1007/ s10489-022-03355-0

    doi

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.