Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

Swin-TUNA : A Novel PEFT Approach for Accurate Food Image Segmentation

T0 review · 3 major / 6 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 A frozen Swin Transformer with 8.13M trainable adapter parameters beats the fully fine-tuned FoodSAM on two food segmentation benchmarks.

desk verdict A sound incremental adapter design with matched PEFT baselines, but the headline claim of beating full fine-tuning is unsupported because the full fine-tuning baselines are protocol-mismatched and even linear probing beats them. read the letter →

arxiv 2507.17347 v3 pith:2SHHPJXB submitted 2025-07-23 cs.CV cs.AI

Haotian Chen , Zhiyong Xiao This is my paper · ORCID
classification cs.CVcs.AI
keywords foodimagesegmentationparameter-efficientfine-tuningSwinTransformeradaptermodulesemantichierarchicalfeatureadaptationSeg103UECFoodPixComplete
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 claims that a frozen Swin Transformer equipped with small per-layer adapter modules—a parameter-efficient fine-tuning (PEFT) approach—can segment food images more accurately than FoodSAM, a fully fine-tuned model with 632.75 million parameters. On the FoodSeg103 benchmark the method reaches 50.56 mIoU, and on UECFoodPix Complete it reaches 74.94 mIoU, both above FoodSAM's published scores, while training only about 4% of the network's parameters (8.13 million). The practical stake is that precise food segmentation for dietary monitoring or industrial inspection currently depends on heavy Transformer models that are costly to train and slow to run. If the claim holds, lightweight deployment and quick adaptation to new food domains become realistic with existing frozen backbones.

What carries the argument

The load-bearing mechanism is the TUNA module, a parallel branch inside each Swin Transformer block consisting of a down-projection, a depthwise convolution, a 1x1 convolution, an up-projection, GeLU activation, and dropout, with a residual connection feeding back into the block. Its named innovation is Hierarchical Feature Adaptation: the depthwise kernel size and the mapping dimension are chosen per stage, with kernel sizes [7,5,5,3] and bottleneck dimensions [64,64,96,192], so shallow and deep layers receive different adapter structures. Two trainable scale parameters, s1 and s2, multiply the frozen task-agnostic features and the adapter's task-specific features, letting the network learn how much of each contributes to the output.

What would settle it

Retrain FoodSAM under Swin-TUNA's exact 640x640 crops on FoodSeg103 and 512x512 crops on UECFoodPix, matching batch size and iteration count, and compare mIoU; if FoodSAM then reaches 50.56% or higher on FoodSeg103, or 74.94% or higher on UECFoodPix Complete, the claim that an 8.13M-parameter adapter surpasses the 632.75M-parameter FoodSAM would fail.

Watch

Extended reading notes

Core claim

On its own terms, the discovery is that a hierarchy-aware PEFT adapter can exceed full fine-tuning on a fine-grained dense prediction task. Swin-TUNA freezes a Swin-L backbone and inserts a TUNA module in parallel with each Swin block, training only the adapters and the segmentation head. With kernel sizes and bottleneck widths that vary across the four stages (kernel sizes [7,5,5,3] and dimensions [64,64,96,192]), the module treats shallow texture features and deep semantic features differently, and two trainable scalars balance the frozen task-agnostic branch against the adapter's task-specific branch. The measured outcome is 50.56 mIoU on FoodSeg103 and 74.94 mIoU on UECFoodPix Complete, ahead of every PEFT baseline tested and, on both datasets, ahead of the fully fine-tuned FoodSAM.

Load-bearing premise

The load-bearing assumption is that the published FoodSAM and FDSNet results are comparable baselines, even though they were produced under different crop sizes and training protocols than Swin-TUNA.

Editorial extensions

If this is right

  • A frozen large backbone can be reused for a new dense prediction task with a small trainable branch, so food segmentation models can run on cheaper hardware with less memory than full fine-tuning.
  • Adapter tuning preserves pre-training knowledge, which the paper argues explains why Swin-TUNA converges faster and degrades less in low-data settings than full fine-tuning.
  • Because Swin-TUNA outperforms AdaptFormer, Mona, BitFit, and linear probing on both benchmarks, the result implies that layer-dependent adapter design matters more than raw trainable parameter count.
  • On FoodSeg103 the method also surpasses FDSNet, a full fine-tuning hybrid baseline, supporting the paper's positioning of PEFT as a path to competitive accuracy rather than a compromise.

Reading between the lines

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

  • A fair head-to-head test would rerun FoodSAM and FDSNet under Swin-TUNA's exact crop sizes, batch sizes, and iteration budgets; the published baselines used 768x768 crops while Swin-TUNA used 640x640 on FoodSeg103 and 512x512 on UECFoodPix, so the reported margins might shrink or invert under matched protocols.
  • The 98.7% parameter reduction compares Swin-TUNA's trainable count with FoodSAM's total count, so a more direct efficiency claim would report total inference-time parameters and FLOPs, where the frozen Swin-L backbone still contributes substantially.
  • The layer-wise recipe of small kernels and narrow bottlenecks in early stages and larger kernels and wider bottlenecks deeper could be tested on other transformer backbones or small-sample segmentation tasks such as medical imaging, since the paper's stated mechanism is not food-specific.
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 / 6 minor

Summary. The paper proposes Swin-TUNA, a parameter-efficient fine-tuning (PEFT) method for food image segmentation. The method inserts tunable adapter modules into a frozen Swin-L backbone, with stage-dependent depthwise convolution kernel sizes and bottleneck dimensions, plus two trainable scaling factors s1 and s2 to balance frozen and adapted features. Experiments on FoodSeg103 and UECFoodPix Complete report mIoU values of 50.56% and 74.94%, respectively, with 8.13M trainable parameters, and the paper claims that this surpasses the fully parameterized FoodSAM model while using only about 4% of the parameters. The paper also reports faster convergence and better low-data generalization than other PEFT baselines, and the code is made publicly available.

Significance. If the central comparison were properly controlled, the result would be practically significant: achieving state-of-the-art food segmentation accuracy with a small fraction of trainable parameters would be a useful contribution, and the hierarchical adapter design is a plausible, easily transferable idea. The paper has real strengths: it provides public code, evaluates on two established benchmarks, and the PEFT baselines in Tables 1 and 2 appear to be matched in crop size and batch size. However, the main advertised claim, that PEFT surpasses full fine-tuning, currently rests on comparisons to published FoodSAM and FDSNet numbers that were obtained under different crop sizes and with different architectures. The significance is therefore conditional on the authors either providing a same-protocol full fine-tuning baseline or reframing the claim.

major comments (3)
  1. [§4.2, Tables 1 and 2] The headline claim that Swin-TUNA surpasses full fine-tuning is not supported by a controlled comparison. FoodSAM and FDSNet are quoted from earlier papers trained at 768×768 crops, whereas Swin-TUNA is trained at 640×640 on FoodSeg103 and 512×512 on UECFoodPix Complete, and those baselines use different backbones and segmentation heads (FoodSAM is SAM-based, not a fine-tuned Swin-L). No full fine-tuning of the paper's own Swin-L/segmentation-head pipeline under the same recipe is reported. This matters because the quoted FoodSAM number on FoodSeg103 (46.42) is already below the paper's own linear probing result (47.85), which strongly suggests protocol mismatch rather than a true advantage of the adapter. I request a same-protocol full fine-tuning baseline, or a clear restatement of the claim as 'outperforms published FoodSAM/FDSNet numbers' rather than 'surpassing full fine-tuning.'
  2. [§4.4, Table 3] The final configuration of Swin-TUNA—kernel sizes [7,5,5,3] and bottleneck widths [64,64,96,192]—is selected by comparing FoodSeg103 test-set mIoU values, and no validation split or model-selection protocol is described. Using the test set for hyperparameter selection can inflate the reported test mIoU relative to an independent evaluation, and it makes it difficult to judge whether the advantage over Mona, AdaptFormer, and BitFit is robust. Please report a validation-based selection procedure, or explicitly evaluate the selected configuration on a held-out test split that was not used for any model choice.
  3. [Tables 1, 2, and Figure 4] No repeated runs, seeds, or confidence intervals are reported. The decisive differences are small—for example, 50.56 vs. 49.39 for BitFit and 49.31 for Mona on FoodSeg103, and 74.94 vs. 75.89 for FDSNet on UECFoodPix—so single-run results cannot establish that the ranking is stable. The convergence claims in Figure 4 are also based on single curves. I request at least multiple seeds (or a statement that these are single runs) and variance estimates for the main comparisons, so that the reader can assess whether the observed gaps are within run-to-run noise.
minor comments (6)
  1. [§4.2] The sentence 'Swin-TUNA outperforms the other PEFT methods on FoodSeg103 and eventually outperforms all other methods' is contradicted by Table 2, where FDSNet reports 75.89 mIoU versus Swin-TUNA's 74.94; please correct this sentence.
  2. [Abstract and §4.2] The abstract and conclusion say that only 4% of parameters are updated, while the text in §4.2 says 'with a mere 8% of the parameters trained'; please check which number is intended and make it consistent.
  3. [Eq. (5)] Equation (5) contains 'oplus' in the running text and the residual term appears to add z^{l-1} twice; the notation should be cleaned up so the reader can follow the exact computation.
  4. [Figure 5(a)] The caption for Figure 5(a) describes both structures as 'parallel'; the right side is described in the text as 'sequential', so the caption should be fixed.
  5. [Table 3] The table headers '%' are unclear; use labels such as 'fixed' vs. 'adaptive' for the two factors instead.
  6. [Section 5] The final sentence, 'providing an efficient solution for assembling lightweight food image,' is incomplete and should be rewritten.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: reported mIoU values are measured benchmark outcomes, not restatements of the model's inputs; baseline crop-size mismatches are a fairness concern, not circularity.

full rationale

The paper contains no formal derivation chain in which a predicted quantity is equal by construction to an input. Equations (4) and (5) define the adapter's residual computation and the trainable scales; s1=1e-6 and s2=0 make the TUNA branch vanish at initialization, so the final 50.56/74.94 mIoU values are empirical outcomes of training on fixed public splits, not identities. The layer-dependent kernel sizes [7,5,5,3] and dimension sizes [64,64,96,192] were selected by ablations on FoodSeg103 (Table 3); this is ordinary hyperparameter/model selection, not a fitted parameter later relabeled as a prediction, and the reported test numbers are measured rather than forced. The only self-citation that appears in the headline comparison chain is FDSNet [34], a prior published baseline co-authored by one of the present authors, but it is not used to justify the method's design and the central 'surpasses FoodSAM' claim rests on the independent FoodSAM numbers; the PEFT baselines (Mona, AdaptFormer, BitFit, linear probing) are run under matched settings in the paper. The different crop sizes between quoted full-fine-tuning baselines (768x768) and Swin-TUNA (640x640 or 512x512) are a legitimate experimental-controls weakness that affects whether 'surpassing full fine-tuning' is fully supported, but it does not make the reported mIoU a restatement of an input. No self-definition, imported uniqueness theorem, ansatz-smuggled-via-citation, or renaming of a known result was identified.

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

The central claim rests on empirical fitting choices, namely the learnable scales and the per-stage kernel and width schedules, and on the comparability of externally quoted baselines. No new physical or ontological entities are introduced.

free parameters (3)
  • Layer scale factors s1 and s2
    Introduced in Eq. 5 and trained by gradient descent for each block. s1 is initialized to 1e-6 and s2 to 0; final values are not reported.
  • Per-stage convolution kernel sizes [7,5,5,3] = [7,5,5,3]
    Chosen by ablation on the FoodSeg103 test set in Section 4.4 Table 3, not predicted from first principles.
  • Per-stage bottleneck dimensions [64,64,96,192] = [64,64,96,192]
    Selected via ablation on the FoodSeg103 test set; this schedule is central to the hierarchical feature adaptation claim.
assumptions (3)
  • domain assumption Swin-L pretrained on ImageNet-22K provides transferable features for food segmentation when frozen
    The whole PEFT strategy assumes the frozen pretrained backbone retains useful features; invoked throughout Section 3 and the experiments.
  • domain assumption Published FoodSAM and FDSNet results are comparable to Swin-TUNA despite different crop sizes, backbones, and training protocols
    Tables 1 and 2 quote full fine-tuning numbers from prior papers without matched reimplementation.
  • domain assumption FoodSeg103 and UECFoodPix Complete annotations are correct and their train and test splits are used as distributed
    Evaluation uses standard benchmarks; no data cleaning or label verification is described.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Swin-TUNA : A Novel PEFT Approach for Accurate Food Image Segmentation." pith.science (2026). https://pith.science/paper/2SHHPJXB

@misc{pith2026250717347,
  author       = {Pith},
  title        = {Pith review of: Swin-TUNA : A Novel PEFT Approach for Accurate Food Image Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2SHHPJXB}},
  note         = {Machine review of arXiv:2507.17347}
}
read the original abstract

In the field of food image processing, efficient semantic segmentation techniques are crucial for industrial applications. However, existing large-scale Transformer-based models (such as FoodSAM) face challenges in meeting practical deploymentrequirements due to their massive parameter counts and high computational resource demands. This paper introduces TUNable Adapter module (Swin-TUNA), a Parameter Efficient Fine-Tuning (PEFT) method that integrates multiscale trainable adapters into the Swin Transformer architecture, achieving high-performance food image segmentation by updating only 4% of the parameters. The core innovation of Swin-TUNA lies in its hierarchical feature adaptation mechanism: it designs separable convolutions in depth and dimensional mappings of varying scales to address the differences in features between shallow and deep networks, combined with a dynamic balancing strategy for tasks-agnostic and task-specific features. Experiments demonstrate that this method achieves mIoU of 50.56% and 74.94% on the FoodSeg103 and UECFoodPix Complete datasets, respectively, surpassing the fully parameterized FoodSAM model while reducing the parameter count by 98.7% (to only 8.13M). Furthermore, Swin-TUNA exhibits faster convergence and stronger generalization capabilities in low-data scenarios, providing an efficient solution for assembling lightweight food image.

Figures

Figures reproduced from arXiv: 2507.17347 by the authors.

Figure 1
Figure 1. Comparison results of Swin-TUNA, FoodSAM, and other PEFT methods in terms of parameter count and [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The results of Swin-TUNA and other PEFT models in segmenting images on FoodSeg103 [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Architectural Overview of TUNA. Left: The proposed Swin-TUNA . Right: Detail of TUNA. TUNA follows the paradigm of Adapter[39] and consists of an upsampling module, a downsampling module, and a non￾linear module (TUNA uses Depthwise convolution). Unlike Adapter, all three modules of TUNA have hierarchical adaptivity, which enables better processing of features at different levels and endows TUNA with the ability to … view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Comparison of Swin-TUNA and other PEFT models on FoodSeg103 and UECFoodPix Complete. The [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: 5(a) Two structures of the proposed method. The left side is the parallel structure and the right side is [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]

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. Food Image Segmentation with LLM-Derived Ingredient Labels and Multimodal Fusion

    cs.CV 2026-07 conditional novelty 6.0 of 10

    Injecting BERT-encoded, LLM-generated ingredient labels into segmentation features and decoder queries raises FoodSeg103 mIoU from 51.9 (Mask2Former baseline) to 54.4 with LIM-F and 55.0 with LIM-Q.

Reference graph

Works this paper leans on

53 extracted references · 42 canonical work pages · cited by 1 Pith paper

  1. [1]

    Application of computer vision techniques to fermented foods: An overview

    Zheli Song, Yuanbo Li, Hongyuan Zhao, Xiaogang Liu, Hailong Ding, Qiansu Ding, Dongna Ma, Shuangping Liu, and Jian Mao. Application of computer vision techniques to fermented foods: An overview. Trends in Food Science & Technology, 160:104982, 2025

    work page 2025

  2. [2]

    Fgfoodnet: Ingredient-perceived fine-grained food recognition for dietary monitoring

    Zhiyong Xiao, Yida Sun, and Zhaohong Deng. Fgfoodnet: Ingredient-perceived fine-grained food recognition for dietary monitoring. Journal of Food Measurement and Characterization, 2025 JUN 28 2025

    work page 2025

  3. [3]

    Fine-grained crop pest classification based on multi-scale feature fusion and mixed attention mechanisms

    Yiheng Qian, Zhiyong Xiao, and Zhaohong Deng. Fine-grained crop pest classification based on multi-scale feature fusion and mixed attention mechanisms. Frontiers in Plant Science, page 1500571, 2025

    work page 2025

  4. [4]

    Swin attention augmented residual network: a fine-grained pest image recognition method

    Xiang Wang, Zhiyong Xiao, and Zhaohong Deng. Swin attention augmented residual network: a fine-grained pest image recognition method. Frontiers in plant science, page 1619551, 2025

    work page 2025

  5. [5]

    Highly scalable parallel genetic algorithm on sunway many-core processors

    Zhiyong Xiao, Xu Liu, Jingheng Xu, Qingxiao Sun, and Lin Gan. Highly scalable parallel genetic algorithm on sunway many-core processors. Future Generation Computer Systems, pages 679–691, 2021

    work page 2021

  6. [6]

    Zhiyong Xiao, Yixin Su, Zhaohong Deng, and Weidong Zhang. Efficient combination of cnn and transformer for dual-teacher uncertainty-guided semi-supervised medical image segmentation.Computer Methods and Programs in Biomedicine, 226:107099, 2022

    work page 2022

  7. [7]

    Light3dhs: A lightweight 3d hippocampus segmen- tation method using multiscale convolution attention and vision transformer

    Zhiyong Xiao, Yuhong Zhang, Zhaohong Deng, and Fei Liu. Light3dhs: A lightweight 3d hippocampus segmen- tation method using multiscale convolution attention and vision transformer. NeuroImage, 292:120608, 2024

    work page 2024

  8. [8]

    FoodSAM: Any Food Segmentation

    Xing Lan, Jiayi Lyu, Hanyu Jiang, Kun Dong, Zehai Niu, Yi Zhang, and Jian Xue. FoodSAM: Any Food Segmentation. IEEE Transactions on Multimedia, 2024. arXiv:2308.05938 [cs]

    work page Pith review arXiv 2024

Show all 53 references
  1. [9]

    Berg, Wan-Yen Lo, Piotr Dollár, and Ross Girshick

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer Whitehead, Alexander C. Berg, Wan-Yen Lo, Piotr Dollár, and Ross Girshick. Segment anything. arXiv:2304.02643, 2023

    2023 arXiv

  2. [10]

    P-tuning v2: Prompt tuning can be comparable to fine-tuning universally across scales and tasks, 2022

    Xiao Liu, Kaixuan Ji, Yicheng Fu, Weng Lam Tam, Zhengxiao Du, Zhilin Yang, and Jie Tang. P-tuning v2: Prompt tuning can be comparable to fine-tuning universally across scales and tasks, 2022

    2022

  3. [11]

    Compacter: Efficient low-rank hypercomplex adapter layers, 2021

    Rabeeh Karimi Mahabadi, James Henderson, and Sebastian Ruder. Compacter: Efficient low-rank hypercomplex adapter layers, 2021

    2021

  4. [12]

    Making pre-trained language models better few-shot learners

    Tianyu Gao, Adam Fisch, and Danqi Chen. Making pre-trained language models better few-shot learners. In Chengqing Zong, Fei Xia, Wenjie Li, and Roberto Navigli, editors, Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th Intern...

    2021

  5. [13]

    Knowledgeable prompt-tuning: Incorporating knowledge into prompt verbalizer for text classification, 2022

    Shengding Hu, Ning Ding, Huadong Wang, Zhiyuan Liu, Jingang Wang, Juanzi Li, Wei Wu, and Maosong Sun. Knowledgeable prompt-tuning: Incorporating knowledge into prompt verbalizer for text classification, 2022

    2022

  6. [14]

    Msp: Multi-stage prompting for making pre-trained language models better translators, 2022

    Zhixing Tan, Xiangwen Zhang, Shuo Wang, and Yang Liu. Msp: Multi-stage prompting for making pre-trained language models better translators, 2022

    2022

  7. [15]

    Mopeft: A mixture-of-pefts for the segment anything model, 2024

    Rajat Sahay and Andreas Savakis. Mopeft: A mixture-of-pefts for the segment anything model, 2024. 11

    2024

  8. [16]

    Swin transformer: Hierarchical vision transformer using shifted windows, 2021

    Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows, 2021

    2021

  9. [17]

    A large-scale benchmark for food image segmentation

    Xiongwei Wu, Xin Fu, Ying Liu, Ee-Peng Lim, Steven CH Hoi, and Qianru Sun. A large-scale benchmark for food image segmentation. In Proceedings of ACM international conference on Multimedia, 2021

    2021

  10. [18]

    UEC-FoodPIX Complete: A large-scale food image segmentation dataset

    Kaimu Okamoto and Keiji Yanai. UEC-FoodPIX Complete: A large-scale food image segmentation dataset. In Proc. of ICPR Workshop on Multimedia Assisted Dietary Management(MADiMa), 2021

    2021

  11. [19]

    An image is worth 16x16 words: Transformers for image recognition at scale, 2021

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at...

    2021

  12. [20]

    Application of improved convolutional neural network in medical image segmentation

    Chen Liu, Zhiyong Xiao, and Nianmao Du. Application of improved convolutional neural network in medical image segmentation. Journal of Frontiers of Computer Science and Technology, pages 1593–1603, 2019

    2019

  13. [21]

    Application of improved convolutional neural network in lung image segmentation

    Baoxin Qian, Zhiyong Xiao, and Wei Song. Application of improved convolutional neural network in lung image segmentation. Journal of Frontiers of Computer Science and Technology, pages 1358–1367, 2020

    2020

  14. [22]

    Sr-net: A sequence offset fusion net and re- fine net for undersampled multislice mr image reconstruction

    Zhiyong Xiao, Nianmao Du, Jianjun Liu, and Weidong Zhang. Sr-net: A sequence offset fusion net and re- fine net for undersampled multislice mr image reconstruction. COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE, 202, 2021

    2021

  15. [23]

    Multi-view hierarchical split network for brain tumor segmentation

    Zhiyong Xiao, Kanghui He, Jianjun Liu, and Weidong Zhang. Multi-view hierarchical split network for brain tumor segmentation. Biomedical Signal Processing and Control, 69, 2021

    2021

  16. [24]

    Tao Liu, Yixin Su, Jiabao Zhang, Tianqi Wei, and Zhiyong Xiao. 3d u-net applied to simple attention module for head and neck tumor segmentation in pet and ct images.Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Note...

    2022

  17. [25]

    Semi-supervised ct image segmentation via contrastive learning based on entropy constraints

    Zhiyong Xiao, Hao Sun, and Fei Liu. Semi-supervised ct image segmentation via contrastive learning based on entropy constraints. Biomedical Engineering Letters, pages 1023–1035, 2024

    2024

  18. [26]

    Pyramid vision transformer: A versatile backbone for dense prediction without convolutions, 2021

    Wenhai Wang, Enze Xie, Xiang Li, Deng-Ping Fan, Kaitao Song, Ding Liang, Tong Lu, Ping Luo, and Ling Shao. Pyramid vision transformer: A versatile backbone for dense prediction without convolutions, 2021

    2021

  19. [27]

    Constrained nonnegative matrix factorization and hyperspectral image dimensionality reduction

    Zhiyong Xiao and Salah Bourennane. Constrained nonnegative matrix factorization and hyperspectral image dimensionality reduction. Remote Sensing Letters, 5(1):46–54, 2014

    2014

  20. [28]

    A spectral–spatial similarity-based method and its application to hyperspectral image classifica- tion

    Zhiyong) Xiao. A spectral–spatial similarity-based method and its application to hyperspectral image classifica- tion. Remote Sensing Letters, pages 122–130, 2014

    2014

  21. [29]

    High accuracy food image classification via vision transformer with data augmentation and feature augmentation

    Xinle Gao, Zhiyong Xiao, and Zhaohong Deng. High accuracy food image classification via vision transformer with data augmentation and feature augmentation. Journal of Food Engineering, 365:111833, 2024

    2024

  22. [30]

    Foodcswin: A high-accuracy food image recognition model for dietary assessment

    Zhiyong Xiao, Ruke Ling, and Zhaohong Deng. Foodcswin: A high-accuracy food image recognition model for dietary assessment. Journal of Food Composition and Analysis, 139:107110, 2025

    2025

  23. [31]

    Fine grained food image recognition based on swin trans- former

    Zhiyong Xiao, Guang Diao, and Zhaohong Deng. Fine grained food image recognition based on swin trans- former. Journal of Food Engineering, 380:112134, 2024

    2024

  24. [32]

    Ovfoodseg: Elevating open-vocabulary food image segmentation via image-informed textual representation, 2024

    Xiongwei Wu, Sicheng Yu, Ee-Peng Lim, and Chong-Wah Ngo. Ovfoodseg: Elevating open-vocabulary food image segmentation via image-informed textual representation, 2024

    2024

  25. [33]

    Canet: cross attention network for food image segmentation

    Xiaoxiao Dong, Haisheng Li, Xiaochuan Wang, Wei Wang, and Junping Du. Canet: cross attention network for food image segmentation. Multimedia Tools and Applications, 83(21):60987–61006, 2024

    2024

  26. [34]

    Food image segmentation based on deep and shallow dual-branch network

    Zhiyong Xiao, Yang Li, and Zhaohong Deng. Food image segmentation based on deep and shallow dual-branch network. Multimedia Systems, 31:85, 2025

    2025

  27. [35]

    Goodfellow, Mehdi Mirza, Da Xiao, Aaron Courville, and Yoshua Bengio

    Ian J. Goodfellow, Mehdi Mirza, Da Xiao, Aaron Courville, and Yoshua Bengio. An empirical investigation of catastrophic forgetting in gradient-based neural networks, 2015

    2015

  28. [36]

    Catastrophic interference in connectionist networks: The sequential learning problem

    Michael McCloskey and Neal J Cohen. Catastrophic interference in connectionist networks: The sequential learning problem. In Psychology of learning and motivation, volume 24, pages 109–165. Elsevier, 1989

    1989

  29. [37]

    Mitigating the alignment tax of rlhf, 2024

    Yong Lin, Hangyu Lin, Wei Xiong, Shizhe Diao, Jianmeng Liu, Jipeng Zhang, Rui Pan, Haoxiang Wang, Wen- bin Hu, Hanning Zhang, Hanze Dong, Renjie Pi, Han Zhao, Nan Jiang, Heng Ji, Yuan Yao, and Tong Zhang. Mitigating the alignment tax of rlhf, 2024

    2024

  30. [38]

    Tsaftaris, and Timothy Hospedales

    Raman Dutt, Linus Ericsson, Pedro Sanchez, Sotirios A. Tsaftaris, and Timothy Hospedales. Parameter-efficient fine-tuning for medical image analysis: The missed opportunity, 2024. 12

    2024

  31. [39]

    Parameter-efficient transfer learning for nlp, 2019

    Neil Houlsby, Andrei Giurgiu, Stanislaw Jastrzebski, Bruna Morrone, Quentin de Laroussilhe, Andrea Ges- mundo, Mona Attariyan, and Sylvain Gelly. Parameter-efficient transfer learning for nlp, 2019

    2019

  32. [40]

    Adapterfusion: Non- destructive task composition for transfer learning, 2021

    Jonas Pfeiffer, Aishwarya Kamath, Andreas Rücklé, Kyunghyun Cho, and Iryna Gurevych. Adapterfusion: Non- destructive task composition for transfer learning, 2021

    2021

  33. [41]

    Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen

    Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. Lora: Low-rank adaptation of large language models, 2021

    2021

  34. [42]

    Bitfit: Simple parameter-efficient fine-tuning for transformer-based masked language-models, 2022

    Elad Ben Zaken, Shauli Ravfogel, and Yoav Goldberg. Bitfit: Simple parameter-efficient fine-tuning for transformer-based masked language-models, 2022

    2022

  35. [43]

    Visual prompt tuning, 2022

    Menglin Jia, Luming Tang, Bor-Chun Chen, Claire Cardie, Serge Belongie, Bharath Hariharan, and Ser-Nam Lim. Visual prompt tuning, 2022

    2022

  36. [44]

    Three things everyone should know about vision transformers, 2022

    Hugo Touvron, Matthieu Cord, Alaaeldin El-Nouby, Jakob Verbeek, and Hervé Jégou. Three things everyone should know about vision transformers, 2022

    2022

  37. [45]

    Adaptformer: Adapting vision transformers for scalable visual recognition

    Shoufa Chen, Chongjian Ge, Zhan Tong, Jiangliu Wang, Yibing Song, Jue Wang, and Ping Luo. Adaptformer: Adapting vision transformers for scalable visual recognition. arXiv preprint arXiv:2205.13535, 2022

    2022 arXiv

  38. [46]

    5%>100%: Breaking performance shackles of full fine-tuning on visual recognition tasks, 2024

    Dongshuo Yin, Leiyi Hu, Bin Li, Youqun Zhang, and Xue Yang. 5%>100%: Breaking performance shackles of full fine-tuning on visual recognition tasks, 2024

    2024

  39. [47]

    Very deep convolutional networks for large-scale image recognition, 2015

    Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale image recognition, 2015

    2015

  40. [48]

    Visualizing and understanding convolutional networks, 2013

    Matthew D Zeiler and Rob Fergus. Visualizing and understanding convolutional networks, 2013

    2013

  41. [49]

    Deep residual learning for image recognition, 2015

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition, 2015

    2015

  42. [50]

    MMSegmentation: Openmmlab semantic segmentation toolbox and benchmark, 2020

    MMSegmentation Contributors. MMSegmentation: Openmmlab semantic segmentation toolbox and benchmark, 2020

    2020

  43. [51]

    Gourmetnet: Food segmentation using multi-scale waterfall features with spatial and channel attention

    Udit Sharma, Bruno Artacho, and Andreas Savakis. Gourmetnet: Food segmentation using multi-scale waterfall features with spatial and channel attention. Sensors, 21(22), 2021

    2021

  44. [52]

    Bayesian deep learning for semantic segmentation of food images

    Eduardo Aguilar, Bhalaji Nagarajan, Beatriz Remeseiro, and Petia Radeva. Bayesian deep learning for semantic segmentation of food images. Computers and Electrical Engineering, 103:108380, 2022

    2022

  45. [53]

    Large scale visual food recognition, 2023

    Weiqing Min, Zhiling Wang, Yuxin Liu, Mengjiang Luo, Liping Kang, Xiaoming Wei, Xiaolin Wei, and Shuqiang Jiang. Large scale visual food recognition, 2023. 13

    2023

Pith tools

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