Pith. sign in

REVIEW 5 major objections 6 minor 1 cited by

MCFNet: A Multimodal Collaborative Fusion Network for Fine-Grained Semantic Classification

T0 review · 5 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read MCFNet reports 93.14% and 92.23% accuracy on the Con-Text and Drink Bottle fine-grained classification benchmarks.

desk verdict A plausible engineering combo whose reported gains are small and whose experimental record is internally inconsistent; the SOTA claim needs a controlled rerun before it can be trusted. read the letter →

arxiv 2505.23365 v1 pith:HQT6JRVY submitted 2025-05-29 cs.CV

Yang Qiao , Xiaoyu Zhong , Xiaofeng Gu , Zhiguo Yu This is my paper · ORCID
classification cs.CV
keywords multimodalfusionfine-grainedclassificationhybridattentionregularizationweightedvotingvision-languagemodelsCon-TextdatasetDrinkBottle
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 proposes MCFNet, a multimodal fusion network for fine-grained classification that combines per-modality regularization (dropout and elastic-net), a hybrid self- and cross-attention mechanism, and a multi-loss decision module with weighted voting. The paper claims that this architecture achieves state-of-the-art accuracy of 93.14% on Con-Text and 92.23% on Drink Bottle, outperforming all listed baselines, including unimodal encoders and recent vision-language models such as Qwen2-VL. The reported ablations attribute the gain to all three modules, with hybrid attention contributing the largest single improvement on Con-Text and the multi-loss module the largest on Drink Bottle. If these results hold, the paper's message is that a lightweight, modular fusion scheme can beat much larger pretrained baselines on close-visual-similarity classification tasks.

What carries the argument

The load-bearing mechanism is the multimodal regularized integrated fusion module paired with the decision classification module. The hybrid attention network runs intra-modal self-attention (average pooling over image patches; multi-scale 1D convolutions over text n-grams) and then inter-modal cross-attention, producing a fused representation $O_H$; in parallel, dropout-regularized and elastic-net-regularized copies of the unimodal features are concatenated and linearly projected into $O_I$ and $O_T$. Three softmax branches yield three losses combined as $Loss = (1-\gamma)Loss_H + \gamma(Loss_T + Loss_I)$ with $\gamma = 0.1$, and the branch probabilities are fused by weighted voting. This composition is what the paper credits for capturing subtle cross-modal semantics while preserving unimodal discriminative information.

What would settle it

Fine-tune Qwen2-VL (or the next-best baseline) on the Con-Text and Drink Bottle splits used in the paper, using the same image size, optimizer, learning rates, and 300-epoch schedule described in Section 4.1, and check whether its reported 92.89% and 91.89% accuracy holds or moves to or beyond MCFNet's 93.14% and 92.23%. A gap that shrinks below noise or reverses would invalidate the state-of-the-art claim.

Watch

Extended reading notes

Core claim

The central claim is that deliberate regularization and multi-level attention alignment in a fusion network yield a consistent edge in fine-grained semantic classification. MCFNet extracts image features with a pretrained Vision Transformer and text features with ALBERT, then applies dropout and elastic-net regularization to each modality, aligns the two modalities through a hybrid attention block (self-attention within each modality followed by cross-attention between them), and combines three classification branches through weighted voting with a composite loss. On the paper's reported experiments, MCFNet reaches 93.14% accuracy and 90.37% F1 on Con-Text, and 92.23% accuracy and 90.02% F1 on Drink Bottle, placing it ahead of the 17 baselines compared. The ablation study shows that removing any of the three modules degrades accuracy on both datasets, and that the hybrid attention module is the most important single component on Con-Text while the multi-loss module matters most on Drink Bottle.

Load-bearing premise

The central comparison is fair: the baselines, especially Qwen2-VL, were evaluated under the same data splits, preprocessing, and training protocol as MCFNet, even though the paper only describes MCFNet's own setup.

Editorial extensions

If this is right

  • If the reported numbers hold, MCFNet outperforms every unimodal and multimodal baseline listed in Table 2 on accuracy, precision, recall, and F1 on both Con-Text and Drink Bottle.
  • The ablation results imply that the hybrid attention module is the strongest single contributor on Con-Text, while the multi-loss module is the strongest on Drink Bottle; removing any module costs accuracy on both datasets.
  • The optimal $\gamma = 0.1$ indicates that auxiliary unimodal losses improve the primary multimodal classification loss when given a small weight, and that setting $\gamma = 0$ (removing multi-loss) hurts performance.
  • Compared with the two attention modules in Table 4, the hybrid attention design is both more accurate (93.14% vs 89.67% and 90.31%) and faster (11.25 ms vs 19.36 ms and 14.38 ms), suggesting the reported efficiency advantage is tied to the specific attention design.

Reading between the lines

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

  • The claimed gains over Qwen2-VL are small (0.25% and 0.34% absolute accuracy), and the paper does not describe the baselines' training protocols; a fair-controlled replication where every baseline is fine-tuned under Section 4.1's exact settings is needed before the state-of-the-art claim can be treated as robust.
  • The text modality in Con-Text and Drink Bottle is essentially OCR-derived scene text; the same architecture could be tested on tasks with richer text (product descriptions, medical reports) to see whether the hybrid attention's n-gram convolutions generalize beyond short phrases.
  • The weighted-voting decision fusion invites a natural extension: replacing the learned fixed weights with confidence-calibrated weights could improve robustness on samples where one modality is noisy or missing.
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

5 major / 6 minor

Summary. The manuscript proposes MCFNet, an ALBERT+ViT based multimodal fusion network for fine-grained classification. The architecture combines a regularized integrated fusion module (dropout, elastic-net regularization, and a hybrid self/cross-attention mechanism), a multimodal decision module with weighted voting, and a multiple-loss training objective. The authors report state-of-the-art accuracy on the Con-Text (93.14%) and Drink Bottle (92.23%) datasets, with ablation studies attributing the gains to each of the three modules. The central claim is empirical: MCFNet outperforms Qwen2-VL by 0.25–0.34 percentage points, and every module contributes positively.

Significance. Fine-grained multimodal classification is an active area, and the proposed combination of modality-specific regularization with hybrid cross-modal attention is a reasonable engineering design. The paper provides a relatively clear architectural description and a broad set of ablations, PR curves, and qualitative examples. However, the empirical claims are not currently verifiable: the baseline protocols are unspecified, the reported margins are very small, and the ablation tables and figures contain internal inconsistencies. If the numerical record is corrected and rerun under controlled conditions, the contribution would be an incremental but potentially useful application of known components. No code or detailed baseline setup is provided, and no parameter-free derivations or machine-checked proofs are involved.

major comments (5)
  1. [§4.3.2, Figure 6, Eq. (15), Table 3] The γ-ablation in Figure 6 reports Con-Text accuracy between roughly 67% and 77% for γ ∈ {0, 0.1, ..., 0.5}, yet Table 3 reports the corresponding full-model configurations at 92–93%. In particular, γ = 0 in Eq. (15) removes MLF, so the γ = 0 point should equal the HAM+RM row of Table 3 (92.83% Con-Text, 91.64% Drink Bottle), but Figure 6 shows about 75–77% and 69–72%. This ~15-point discrepancy cannot be explained by metric choice or minor implementation differences; at least one of the two experimental records is incorrect, which directly undermines the claim that MLF contributes the reported gains and that the final accuracy is 93.14%.
  2. [§4.1.2, Table 2] The baseline protocol is not specified. Section 4.1.2 gives training details for MCFNet only; there is no description of how Flamingo, LLaVA, Qwen2-VL, TomBERT, ViT-BERT, or the other baselines were adapted to these datasets, which splits were used for them, what prompts or input formats were employed, or how their hyperparameters were chosen. Because the reported gains over Qwen2-VL are only 0.25–0.34 percentage points, an uncontrolled or zero-shot baseline evaluation would erase the state-of-the-art claim. The authors should report the exact baseline setup, ideally with the same training code and splits, or restrict the claims to comparisons that were actually controlled.
  3. [§3.3.1, Eq. (6)] The ElasticNet operation in Eq. (6) is not a well-defined neural network layer as written. It is an argmin over x′ of a sum involving an unspecified Loss(x, x′), with no closed-form solution, no iterative solver, and no demonstration of differentiability through the argmin; the paper also does not state how α and β are set. This makes the proposed architecture irreproducible as described. The authors should define a concrete differentiable implementation, for example a proximal operator or a fixed-point iteration, and specify how gradients flow through it.
  4. [Table 3 vs. Table 5] The regularization ablation in Table 5 is inconsistent with the component ablation in Table 3. Table 5 reports that removing the hybrid attention module drops Con-Text accuracy to 88.19% and Drink Bottle accuracy to 87.66%, whereas Table 3’s configuration without HAM but with RM and MLF (row "- ✓ ✓") achieves 92.77% and 91.47%, respectively. The 4.58-point (Con-Text) and 3.81-point (Drink Bottle) gaps indicate that different configurations are being compared under the same name, so the reader cannot determine the actual contribution of the hybrid attention module.
  5. [§4.1, Table 2] No error bars, confidence intervals, or multiple-seed results are reported. Given that the headline margins over the second-best model are 0.25% (Con-Text) and 0.34% (Drink Bottle), run-to-run variance could easily change the ranking. The authors should report the mean and standard deviation over at least three independent runs for MCFNet and the strongest baselines.
minor comments (6)
  1. [§1, §3.3.2, §3.4] The text contains unresolved placeholders "Fig.??" in several places, including Section 1, Section 3.3.2, and Section 3.4; these must be replaced with actual figure references.
  2. [Table 2] The table lists "RestNet50" and "LLaVa"; these should be corrected to "ResNet50" and "LLaVA" for consistency with the literature.
  3. [Table 4] The header arrows (Parameter ↑, Time ↓, Accuracy ↓) are confusing; the authors should state explicitly that higher parameter counts are not necessarily better and should clarify whether lower time and higher accuracy are the desired directions.
  4. [§4.1.2] The sentence "the quantization step sizes for both the ImageNet and Drink Bottle datasets are set to 256" appears to refer to input image resizing rather than quantization; this should be reworded to avoid ambiguity.
  5. [Data Availability] The Data Availability statement says data will be made available on request, but no code is mentioned; releasing the training and evaluation code would substantially improve reproducibility.
  6. [§4.3.2, Eq. (15)] The discussion of γ = 0 as equivalent to removing MLF is consistent with Eq. (15), but the text should explicitly clarify that Figure 6 includes this degenerate case and that the statement "when MLF is employed (γ > 0)" excludes the γ = 0 point shown in the same figure.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: MCFNet's accuracy claims are empirical measurements against external benchmarks; no prediction reduces by construction to a fitted input or self-citation.

full rationale

The paper's central claims are experimental: MCFNet achieves 93.14% on Con-Text and 92.23% on Drink Bottle, compared with external baselines, with ablations attributing gains to HAM, RM, and MLF. These are measured outcomes, not quantities derived from fitted parameters by construction. The only hyperparameter analyzed, γ in Eq. (15), is tuned on validation sets, which is standard practice and does not make the reported test accuracy a fitted prediction. The architecture combines standard components (ALBERT, ViT, dropout, elastic-net regularization, attention, multiple losses) without invoking any uniqueness theorem or author-imported constraint that would force the result. The paper does not rely on self-citations for its load-bearing results; cited prior work provides models, datasets, and baselines, not the claimed superiority of MCFNet. The noted inconsistency between Figure 6's γ-ablation accuracy values (~67–77%) and Table 3's corresponding configurations (~92–93%) is a serious empirical verification concern, but it is not a circularity: it concerns whether the reported numbers are reproducible, not whether any equation is equivalent to its inputs by definition. Therefore, no circular step satisfying the quoted-evidence requirement is present, and the appropriate score is 0.

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

The paper's central claim is an empirical accuracy improvement, so the ledger lists the hyperparameters the result depends on (gamma, dropout p, elastic net coefficients) and the assumptions that the benchmarks and baselines are comparable and that the reported numbers are reliable. No new physical or mathematical entities are introduced.

free parameters (3)
  • gamma (loss weighting) = 0.1 (chosen by ablation)
    Eq. (15) combines primary and auxiliary losses; Section 4.3.2 selects 0.1 on the validation sets, and the claimed MLF benefit depends on this tuning.
  • dropout keep probability p = not reported
    Eq. (1)-(2) apply dropout to image and text features, but no value for p is given anywhere, making the regularization strength indeterminate.
  • elastic net coefficients alpha and beta = not reported
    Eq. (4)-(6) define an 'ElasticNet' operation with L1 and L2 coefficients, but alpha and beta are never specified or ablated.
assumptions (4)
  • domain assumption Pre-trained ALBERT and ViT features capture complementary fine-grained cues for the Con-Text and Drink Bottle datasets.
    The whole method assumes that image and text (OCR-like labels) are complementary and that fusing them improves classification; this is asserted in Section 1 and not formally proven.
  • domain assumption The benchmark comparisons are fair, meaning all baselines are trained and evaluated under the same protocol as MCFNet.
    Section 4.1 describes MCFNet's setup but does not describe the protocol used for Flamingo, LLaVA, Qwen2-VL, or other baselines; the SOTA claim presumes equal footing.
  • ad hoc to paper The optimization defined in Eq. (6) is differentiable and trainable as a network layer.
    ElasticNet is written as an arg min over x' with a reconstruction loss plus L1/L2 penalties, but no solver or differentiable approximation is specified; the paper assumes this can be used in end-to-end training.
  • domain assumption The reported numerical results are accurate and not cherry-picked across runs.
    No error bars or multiple-seed statistics are provided, so any claim of consistency across runs rests on this assumption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MCFNet: A Multimodal Collaborative Fusion Network for Fine-Grained Semantic Classification." pith.science (2026). https://pith.science/paper/HQT6JRVY

@misc{pith2026250523365,
  author       = {Pith},
  title        = {Pith review of: MCFNet: A Multimodal Collaborative Fusion Network for Fine-Grained Semantic Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HQT6JRVY}},
  note         = {Machine review of arXiv:2505.23365}
}
read the original abstract

Multimodal information processing has become increasingly important for enhancing image classification performance. However, the intricate and implicit dependencies across different modalities often hinder conventional methods from effectively capturing fine-grained semantic interactions, thereby limiting their applicability in high-precision classification tasks. To address this issue, we propose a novel Multimodal Collaborative Fusion Network (MCFNet) designed for fine-grained classification. The proposed MCFNet architecture incorporates a regularized integrated fusion module that improves intra-modal feature representation through modality-specific regularization strategies, while facilitating precise semantic alignment via a hybrid attention mechanism. Additionally, we introduce a multimodal decision classification module, which jointly exploits inter-modal correlations and unimodal discriminative features by integrating multiple loss functions within a weighted voting paradigm. Extensive experiments and ablation studies on benchmark datasets demonstrate that the proposed MCFNet framework achieves consistent improvements in classification accuracy, confirming its effectiveness in modeling subtle cross-modal semantics.

Figures

Figures reproduced from arXiv: 2505.23365 by the authors.

Figure 2
Figure 2. Detailed network architecture. The network employs pre-trained models ALBERT and ViT to extract textual and visual information, respectively. Subsequently, a multimodal regularization integration fusion network is used to fuse the features of scene text regions with those of visually salient objects. In the multimodal decision classification network, the network utilizes multiple loss functions and incorporates a dy… view at source ↗
Figure 5
Figure 5. PR curves of different ablation methods. 0 0.1 0.2 0.3 0.4 0.5 65 70 75 80 A c c u r a c y ( % ) g Con-Text Drink Bottle 77.16 72.05 75.45 69.38 76.08 68.93 75.23 68.89 75.88 69.67 74.17 67.78 (a) Comparison of accuracy with different 𝛾 values on the Con-Text and Drink Bottle datasets. 0 0.1 0.2 0.3 0.4 0.5 65 70 75 F 1 ( % ) g Con-Text Drink Bottle 72.53 70.14 70.74 65.81 72.14 67.53 70.67 66.80 71.36 67.69 69.88 6… view at source ↗
Figure 6
Figure 6. Experimental results corresponding to different values of the trade off hyperparameter 𝛾 . TL: Dessert Shop Dessert S: 0.98 Barber: 0.34 Ice Cream: 1.7e-2 TL: Chinese R. Chinese R.: 0.99 Dinner: 3.5e-7 Groggery: 8.4e-8 TL: Pizzeria Pizzeria: 0.99 Dinner: 4.6e-5 Restaur.: 3.1e-5 TL: School School: 0.99 Theatre: 1.8e-6 Library: 2.4e-7 TL: Cafe Cafe: 0.98 Dessert S.: 1.3e-3 Bakery: 5.4e-3 TL: Bakery Bakery: 0.96 Drinks… view at source ↗
Figures from the paper (1 more)
Figure 7
Figure 7. Figure 7: Classification prediction. The figure illustrates the true label of a category, along with the top three probability-ranked results from the network’s classification of this category. In this representation, blue is used to indicate correct predictions. Y. Qiao et al.:…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Localization-Infused Vision-Language Semantic Fusion for Text-Guided Medical Image Segmentation

    cs.CV 2026-07 conditional novelty 6.0 of 10

    A localization-infused vision-language fusion method converts textual location cues into multi-scale localization predictions and uses them to guide medical image segmentation, outperforming prior methods on three benchmarks.

Reference graph

Works this paper leans on

61 extracted references · 55 canonical work pages · cited by 1 Pith paper

  1. [1]

    Tuia , author B

    author D. Tuia , author B. Kellenberger , author S. Beery , author B. R. Costelloe , author S. Zuffi , author B. Risse , author A. Mathis , author M. W. Mathis , author F. Van Langevelde , author T. Burghardt , et al., title Perspectives in machine learning for wildlife conservation , journal Nature communications. volume 13 ( year 2022 ) pages 792--806

    work page 2022

  2. [2]

    Fu , author A

    author Z. Fu , author A. Wu , author S. Yang , author T. Ma , author L. He , title Cafnet: Context aligned fusion for depth completion , journal Computer Vision and Image Understanding volume 249 ( year 2024 ) pages 104158

    work page 2024

  3. [3]

    Rubin , author O

    author M. Rubin , author O. Stein , author N. A. Turko , author Y. Nygate , author D. Roitshtain , author L. Karako , author I. Barnea , author R. Giryes , author N. T. Shaked , title TOP-GAN : Stain-free cancer cell classification using deep learning with a small training set , journal Medical image analysis. volume 57 ( year 2019 ) pages 176--185

    work page 2019

  4. [4]

    author Z. C. Chen , author H. X. Zheng , author L. J. Wu , author J. C. Huang , author Y. Yang , title Deep transfer learning based intelligent gunshot detection and firearm recognition using tri-axial acceleration , journal IEEE Internet of Things Journal. volume 12 ( year 2024 ) pages 5891--5900

    work page 2024

  5. [5]

    Prakash , author K

    author A. Prakash , author K. Chitta , author A. Geiger , title Multi-modal fusion transformer for end-to-end autonomous driving , in: booktitle Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , year 2021 , pp. pages 7077--7087

    work page 2021

  6. [6]

    Xu , author Y

    author Q. Xu , author Y. M. Mei , author J. P. Liu , author C. L. Li , title Multimodal cross-layer bilinear pooling for rgbt tracking , journal IEEE Transactions on Multimedia. volume 24 ( year 2021 ) pages 567--580

    work page 2021

  7. [7]

    author J. M. Fu , author S. Y. Xu , author H. D. Liu , author Y. Liu , author N. Xie , author C. C. Wang , author J. Liu , author Y. Sun , author B. Wang , title CMA-CLIP : Cross-modality attention clip for text-image classification , in: booktitle 2022 IEEE international conference on image processing (ICIP) , organization IEEE , year 2022 , pp. pages 2846--2850

    work page 2022

  8. [8]

    author D. G. Low , title Distinctive image features from scale-invariant keypoints , journal Journal of Computer Vision. volume 60 ( year 2004 ) pages 91--110

    work page 2004

Show all 61 references
  1. [9]

    He , author J

    author N. He , author J. H. Cao , author L. Song , title Scale space histogram of oriented gradients for human detection , in: booktitle 2008 International Symposium on Information Science and Engineering , year 2008 , pp. pages 167--170

    2008

  2. [10]

    Xu , author X

    author J. Xu , author X. Q. Zhang , author C. M. Zhao , author Z. L. Geng , author Y. R. Feng , author K. Miao , author Y. J. Li , title Improving fine-grained image classification with multimodal information , journal IEEE Transactions on Multimedia. volume 26 ( year 2023 ) p...

    2023

  3. [11]

    Li , author M

    author K. Li , author M. Huang , author X. Y. Yu , author C. Yang , title Research on F ine-grained visual classification method based on dual-attention feature complementation , journal IEEE Access. volume 12 ( year 2024 ) pages 192209--192218

    2024

  4. [12]

    author R. Y. Ji , author J. Y. Li , author L. B. Zhang , author J. Liu , author Y. J. Wu , title Dual transformer with multi-grained assembly for fine-grained visual classification , journal IEEE Transactions on Circuits and Systems for Video Technology. volume 33 ( year 2023 ...

    2023

  5. [13]

    Huang , author S

    author Z. Huang , author S. Yang , author J. Wu , author L. Zhu , author J. Liu , title Fusiondiff: A unified image fusion network based on diffusion probabilistic models , journal Computer Vision and Image Understanding volume 244 ( year 2024 ) pages 104011

    2024

  6. [14]

    Zhou , author G

    author J. Zhou , author G. Q. Cui , author S. D. Hu , author Z. Y. Zhang , author C. Yang , author Z. Y. Liu , author L. F. Wang , author C. C. Li , author M. S. Sun , title Graph neural networks: A review of methods and applications , journal AI open. volume 1 ( year 2020 ) p...

    2020

  7. [15]

    Huang , author Z

    author G. Huang , author Z. Liu , author L. Van Der Maaten , author K. Q. Weinberger , title Densely connected convolutional networks , in: booktitle Proceedings of the IEEE conference on computer vision and pattern recognition , year 2017 , pp. pages 4700--4708

    2017

  8. [16]

    author J. X. Li , title Fine-grained sentiment analysis with a fine-tuned BERT and an improved pre-training BERT , in: booktitle 2023 IEEE International Conference on Image Processing and Computer Applications (ICIPCA) , organization IEEE , year 2023 , pp. pages 1031--1034

    2023

  9. [17]

    author M. T. Yang , author S. K. Ng , author J. L. Fu , title Omnidialog: An omnipotent pre-training model for task-oriented dialogue system , journal IEEE Transactions on Audio, Speech and Language Processing. volume 1 ( year 2025 ) pages 1--11

    2025

  10. [18]

    author J. L. Fu , author H. L. Zheng , author T. Mei , title Look closer to see better: Recurrent attention convolutional neural network for fine-grained image recognition , in: booktitle Proceedings of the IEEE conference on computer vision and pattern recognition , year 2017...

    2017

  11. [19]

    Sun , author Y

    author M. Sun , author Y. Yuan , author F. Zhou , author E. Ding , title Multi-attention multi-class constraint for fine-grained image recognition , in: booktitle Proceedings of the european conference on computer vision (ECCV) , year 2018 , pp. pages 805--821

    2018

  12. [20]

    author H. L. Zheng , author J. L. Fu , author Z. J. Zha , author J. B. Luo , title Looking for the devil in the details: Learning trilinear attention sampling network for fine-grained image recognition , in: booktitle Proceedings of the IEEE/CVF conference on computer vision a...

    2019

  13. [21]

    author T. Y. Lin , author A. RoyChowdhury , author S. Maji , title Bilinear CNN models for fine-grained visual recognition , in: booktitle Proceedings of the IEEE international conference on computer vision , year 2015 , pp. pages 1449--1457

    2015

  14. [22]

    Gao , author O

    author Y. Gao , author O. Beijbom , author N. Zhang , author T. Darrell , title Compact bilinear pooling , in: booktitle Proceedings of the IEEE conference on computer vision and pattern recognition , year 2016 , pp. pages 317--326

    2016

  15. [23]

    Kong , author C

    author S. Kong , author C. Fowlkes , title Low-rank bilinear pooling for fine-grained classification , in: booktitle Proceedings of the IEEE conference on computer vision and pattern recognition , year 2017 , pp. pages 365--374

    2017

  16. [24]

    Cui , author F

    author Y. Cui , author F. Zhou , author J. Wang , author X. Liu , author Y. Q. Lin , author S. Belongie , title Kernel pooling for convolutional neural networks , in: booktitle Proceedings of the IEEE conference on computer vision and pattern recognition , year 2017 , pp. page...

    2017

  17. [25]

    author D. L. Chang , author Y. F. Ding , author J. Y. Xie , author A. K. Bhunia , author X. X. Li , author Z. Y. Ma , author M. Wu , author J. Guo , author Y. Z. Song , title The devil is in the channels: Mutual-channel loss for fine-grained image classification , journal IEEE...

    2020

  18. [26]

    author Y. M. Rao , author G. Y. Chen , author J. W. Lu , author J. Zhou , title Counterfactual attention learning for fine-grained visual categorization and re-identification , in: booktitle Proceedings of the IEEE/CVF international conference on computer vision , year 2021 , ...

    2021

  19. [27]

    Shen , author J

    author F. Shen , author J. Yu , author C. Wang , author X. Jiang , author X. Du , author J. Tang , title Imaggarment-1: Fine-grained garment generation for controllable fashion design , journal arXiv preprint arXiv:2504.13176 ( year 2025 )

    2025 arXiv

  20. [28]

    Dosovitskiy , author L

    author A. Dosovitskiy , author L. Beyer , author A. Kolesnikov , author D. Weissenborn , author X. H. Zhai , author T. Unterthiner , author M. Dehghani , author M. Minderer , author G. Heigold , author S. Gelly , et al., title An image is worth 16 16 words: transformers for im...

    2020

  21. [29]

    Liu , author Y

    author Z. Liu , author Y. T. Lin , author Y. Cao , author H. Hu , author Y. X. Wei , author Z. Zhang , author S. Lin , author B. N. Guo , title Swin transformer: Hierarchical vision transformer using shifted windows , in: booktitle Proceedings of the IEEE/CVF international con...

    2021

  22. [30]

    He , author J

    author J. He , author J. N. Chen , author S. Liu , author A. Kortylewski , author C. Yang , author Y. Bai , author C. Wang , title Transfg: A transformer architecture for fine-grained recognition , in: booktitle Proceedings of the AAAI conference on artificial intelligence , y...

    2022

  23. [31]

    author Y. Q. Hu , author X. Jin , author Y. Zhang , author H. W. Hong , author J. F. Zhang , author Y. He , author H. Xue , title Rams-trans: Recurrent attention multi-scale transformer for fine-grained image recognition , in: booktitle Proceedings of the 29th ACM internationa...

    2021

  24. [32]

    author X. H. Yu , author J. Wang , author Y. Zhao , author Y. S. Gao , title Mix-ViT : Mixing attentive vision transformer for ultra-fine-grained visual categorization , journal Pattern Recognition. volume 135 ( year 2023 ) pages 109131--109140

    2023

  25. [33]

    Shen , author J

    author F. Shen , author J. Tang , title Imagpose: A unified conditional framework for pose-guided person generation , journal Advances in neural information processing systems volume 37 ( year 2024 ) pages 6246--6266

    2024

  26. [34]

    author H. X. Huang , author J. J. Zhang , author L. T. Yu , author J. Zhang , author Q. Wu , author C. Xu , title TOAN : Target-oriented alignment network for fine-grained image categorization with few labeled samples , journal IEEE Transactions on Circuits and Systems for Vid...

    2021

  27. [35]

    Wang , author J

    author H. Wang , author J. C. Liao , author T. H. Cheng , author Z. W. Gao , author H. Liu , author B. Ren , author X. Bai , author W. Y. Liu , title Knowledge mining with scene text for fine-grained recognition , in: booktitle Proceedings of the IEEE/CVF conference on compute...

    2022

  28. [36]

    Jiang , author M

    author D. Jiang , author M. Ye , title Cross-modal implicit relation reasoning and aligning for text-to-image person retrieval , in: booktitle Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , year 2023 , pp. pages 2787--2797

    2023

  29. [37]

    author S. Y. Xuan , author Q. P. Guo , author M. Yang , author S. L. Zhang , title Pink: Unveiling the power of referential comprehension for multi-modal LLMS , in: booktitle Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , year 2024 , pp. pa...

    2024

  30. [38]

    He , author G

    author H. He , author G. Li , author Z. J. Geng , author J. L. Xu , author Y. X. Peng , title Analyzing and boosting the power of fine-grained visual recognition for multi-modal large language models , journal arXiv preprint arXiv:2501.15140 ( year 2025 )

    2025 arXiv

  31. [39]

    Shen , author X

    author F. Shen , author X. Jiang , author X. He , author H. Ye , author C. Wang , author X. Du , author Z. Li , author J. Tang , title Imagdressing-v1: Customizable virtual dressing , in: booktitle Proceedings of the AAAI Conference on Artificial Intelligence , volume volume 3...

    2025

  32. [40]

    Shen , author C

    author F. Shen , author C. Wang , author J. Gao , author Q. Guo , author J. Dang , author J. Tang , author T.-S. Chua , title Long-term talkingface generation via motion-prior conditional diffusion model , journal arXiv preprint arXiv:2502.09533 ( year 2025 b )

    2025 arXiv

  33. [41]

    Shen , author H

    author F. Shen , author H. Ye , author S. Liu , author J. Zhang , author C. Wang , author X. Han , author Y. Wei , title Boosting consistency in story visualization with rich-contextual conditional diffusion models , in: booktitle Proceedings of the AAAI Conference on Artifici...

    2025

  34. [42]

    Shen , author H

    author F. Shen , author H. Ye , author J. Zhang , author C. Wang , author X. Han , author W. Yang , title Advancing pose-guided image synthesis with progressive conditional diffusion models , journal arXiv preprint arXiv:2310.06313 ( year 2023 )

    arXiv

  35. [43]

    author Z. Z. Lan , author M. D. Chen , author S. Goodman , author K. Gimpel , author P. Sharma , author R. Soricut , title ALBERT : A lite BERT for self-supervised learning of language representations , in: booktitle 8th International conference on learning representations , year 2020

    2020

  36. [44]

    author L. A. Hendricks , author J. Mellor , author R. Schneider , author J.-B. Alayrac , author A. Nematzadeh , title Decoupling the role of data, attention, and losses in multimodal transformers , journal Transactions of the Association for Computational Linguistics volume 9 ...

    2021

  37. [45]

    Karaoglu , author R

    author S. Karaoglu , author R. Tao , author J. C. van Gemert , author T. Gevers , title Con-text: Text detection for fine-grained object classification , journal IEEE transactions on image processing. volume 26 ( year 2017 ) pages 3965--3980

    2017

  38. [46]

    Bai , author M

    author X. Bai , author M. K. Yang , author P. Lyu , author Y. C. Xu , author J. B. Luo , title Integrating scene text and visual appearance for fine-grained image classification , journal IEEE Access. volume 6 ( year 2018 ) pages 66322--66335

    2018

  39. [47]

    Howard , author S

    author J. Howard , author S. Ruder , title Universal language model fine-tuning for text classification , in: booktitle Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics , year 2018 , pp. pages 328--339

    2018

  40. [48]

    author K. M. He , author X. Y. Zhang , author S. Q. Ren , author J. Sun , title Deep residual learning for image recognition , in: booktitle Proceedings of the IEEE conference on computer vision and pattern recognition , year 2016 , pp. pages 770--778

    2016

  41. [49]

    Pi , author W

    author Q. Pi , author W. J. Bian , author G. R. Zhou , author X. Q. Zhu , author K. Gai , title Practice on long sequential user behavior modeling for click-through rate prediction , in: booktitle Proceedings of the 25th ACM SIGKDD international conference on knowledge discove...

    2019

  42. [50]

    author J. F. Yu , author J. Jiang , title Adapting BERT for target-oriented multimodal sentiment classification , in: booktitle Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence , year 2019 , pp. pages 5408--5414

    2019

  43. [51]

    Li , author B

    author Q. Li , author B. Q. Gong , author Y. Cui , author D. Kondratyuk , author X. Z. Du , author M. H. Yang , author M. Brown , title Towards a unified foundation model: Jointly pre-training transformers on unpaired images and text , journal arXiv preprint arXiv:2112.07074 (...

    2021 arXiv

  44. [52]

    author J. B. Alayrac , author J. Donahue , author P. Luc , author A. Miech , author I. Barr , author Y. Hasson , author K. Lenc , author A. Mensch , author K. Millican , author M. Reynolds , et al., title Flamingo: a visual language model for few-shot learning , journal Advanc...

    2022

  45. [53]

    author H. T. Liu , author C. Y. Li , author Q. Y. Wu , author Y. J. Lee , title Visual instruction tuning , journal Advances in neural information processing systems. volume 36 ( year 2023 ) pages 34892--34916

    2023

  46. [54]

    Wang , author S

    author P. Wang , author S. Bai , author S. Tan , author S. J. Wang , author Z. H. Fan , author J. Z. Bai , author K. Q. Chen , author X. J. Liu , author J. L. Wang , author W. B. Ge , et al., title Qwen2-VL : Enhancing vision-language model's perception of the world at any res...

    2024 arXiv

  47. [55]

    Zhang , author D

    author S. Zhang , author D. Q. Zheng , author X. C. Hu , author M. Yang , title Bidirectional long short-term memory networks for relation classification , in: booktitle Proceedings of the 29th Pacific Asia conference on language, information and computation , year 2015 , pp. ...

    2015

  48. [56]

    Tay , author L

    author Y. Tay , author L. A. Tuan , author S. C. Hui , author J. Su , title Reasoning with sarcasm by reading in-between , in: booktitle Proceedings of the 56th annual meeting of the association for computational linguistics , year 2018 , pp. pages 1010--1020

    2018

  49. [57]

    author H. W. Chung , author L. Hou , author S. Longpre , author B. Zoph , author Y. Tay , author W. Fedus , author Y. X. Li , author X. Z. Wang , author M. Dehghani , author S. Brahma , et al., title Scaling instruction-finetuned language models , journal Journal of Machine Le...

    2024

  50. [58]

    author P. Q. Zhuang , author Y. L. Wang , author Y. Qiao , title Learning attentive pairwise interaction for fine-grained classification , in: booktitle Proceedings of the AAAI conference on artificial intelligence , year 2020 , pp. pages 13130--13137

    2020

  51. [59]

    Brock , author S

    author A. Brock , author S. De , author S. L. Smith , author K. Simonyan , title High-performance large-scale image recognition without normalization , in: booktitle Proceedings of the 38th international conference on machine learning , year 2021 , pp. pages 1059--1071

    2021

  52. [60]

    Schifanella , author P

    author R. Schifanella , author P. De Juan , author J. Tetreault , author L. L. Cao , title Detecting sarcasm in multimodal social platforms , in: booktitle Proceedings of the 24th ACM international conference on Multimedia , year 2016 , pp. pages 1136--1145

    2016

  53. [61]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.