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

Jui-Feng Chi; Sheng-Long Lin; Wei-Ta Chu

REVIEW 3 major objections 4 minor 25 references

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

T0 review · 3 major / 4 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read Injecting LLM ingredient labels into Mask2Former reaches 55.0 mIoU on FoodSeg103

desk verdict Clean idea and a real confound: the LLM sees the test image and leaks a per-image ingredient prior, so the mIoU gains are not cleanly attributable to the LIM modules. read the letter →

arxiv 2607.25820 v1 pith:HMQUKZ5X submitted 2026-07-28 cs.CV

Jui-Feng Chi , Wei-Ta Chu , Sheng-Long Lin This is my paper · ORCID

classification cs.CV

keywords foodimagesegmentationingredientlabelslargelanguagemodelsmultimodalfusionMask2FormerLIM-FLIM-QSeg103

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

The paper proposes two lightweight modules, LIM-F and LIM-Q, that inject ingredient labels generated by a large language model into the visual segmentation pipeline. On the FoodSeg103 benchmark, the best configuration reaches 55.0 mIoU, surpassing prior state-of-the-art results by a noticeable margin. Crucially, the method works without paired image-text pretraining or curated recipe corpora, relying on LLM-generated labels from the input image itself. If correct, this suggests that semantic text cues can be effectively plugged into standard segmentation architectures at a modest training-memory cost.

What carries the argument

The two modules are the central mechanism: LIM-F (Language Injection Module for Features) applies cross-attention at each stage of a multi-level encoder, using visual features as queries and BERT-encoded ingredient embeddings as keys and values, with residual connections and feed-forward networks. LIM-Q (Language Injection Module for Queries) operates inside a Transformer decoder, performing cross-attention between learnable query tokens and BERT-encoded ingredient labels at the start of every decoding round, followed by self-attention and a feed-forward network. Both avoid pre-aligned image-text pairs; the ingredient labels are produced per image by an LLM prompted with the image and the fu

What would settle it

Train the same Mask2Former backbone with the LLM ingredient labels provided as a simple per-image class-prior weighting of the classification loss, or as input tokens appended to queries, but without the LIM cross-attention structure. If this simpler conditioning matches the 54.4–55.0 mIoU, the module-level claim collapses. Conversely, replacing the LLM labels with random labels should drop performance if the modules genuinely use semantic content.

Watch

Extended reading notes

Core claim

The authors claim that LLM-derived ingredient labels, encoded by BERT and fused via two simple cross-attention modules, substantially improve fine-grained food segmentation. LIM-Q, injected into the Mask2Former decoder before each round of query updates, yields 55.0 mIoU on FoodSeg103 with a Swin-L backbone, compared with 51.9 for the baseline Mask2Former and 50.6 for the previous best method, Swin-TUNA. LIM-F, which fuses text into multi-level visual features, achieves 54.4 mIoU under the same backbone and also improves a CNN-based K-Net pipeline from 47.7 to 49.0 mIoU. The authors attribute the gains to improved discrimination of visually similar and rare ingredients, and they show the ext

Load-bearing premise

The evaluation assumes that LLM-generated ingredient labels for the test image are a fair system input, and that the mIoU increase over baselines without those labels is attributable to the LIM modules, rather than to the label information alone or to a per-image class prior.

Editorial extensions

If this is right

The plug-and-play design means ingredient semantics can be added to existing segmentation architectures without changing backbones or annotation formats.
Gains are largest for rare and visually ambiguous ingredients, which matters for nutrition tracking where minor ingredients are easily missed.
Eliminating paired image-text pretraining lowers the barrier for applying language-guided segmentation to other domains with scarce aligned data.
The modest memory increase (at most 3.8 GB) makes the approach feasible on a single GPU for training.

Reading between the lines

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

LLM label quality becomes a hidden variable: if the LLM mislabels an ingredient, the module may reinforce an error; the paper does not assess this failure mode.
A class-prior baseline that simply up-weights rare classes might capture part of the same gains, since the LLM labels effectively provide a per-image class prior.
The approach could be extended to open-vocabulary food segmentation if the LLM produces labels beyond the fixed 104 classes.
The faster convergence of LIM-Q relative to LIM-F suggests query-level injection is a more direct channel for semantic guidance; future work might explore even earlier or hierarchical injection.

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.

Desk Editor's Note

read the letter

Two things you should know. First, the idea is actually practical: instead of curating image-text pairs, the authors use an LLM to generate ingredient labels from the input food image and inject them into a segmentation model via two lightweight cross-attention modules (LIM-F and LIM-Q). That removes a real bottleneck and the modules are clearly described. Second, the evaluation is confounded in a way that matters. In Section 4.1, the LLM receives the test image plus the full list of 104 FoodSeg103 classes and is asked to infer which ingredients are present. So every LIM variant gets a per-image semantic prior—essentially a class-presence hint—that the Mask2Former, K-Net, BEiT, and Swin-TUNA baselines do not receive. The reported gains, especially the rare-class jump for 'date' from 1.28 to 88.93 in Table 2, look exactly like what you'd expect from a class-presence prior rather than from the specific cross-attention design. The paper never runs the obvious control: inject the same LLM label list through a trivial mechanism (class-prior bias, label-conditioned classifier head, or simple query bias) and see if the gains persist.

That confound is the main soft spot, and it is load-bearing. The modules may be perfectly good, but as written the attribution of the 51.9 to 55.0 improvement to LIM-Q is not established. I'd also flag the internal inconsistency: the abstract claims 49.8 mIoU for the CNN-based LIM-F while Table 1 reports 49.0. And there are no error bars, no multiple seeds, and no code. Those are secondary, but they make the numbers harder to trust.

Now the credit. The work does more than cherry-pick results: it compares three integration variants, shows convergence curves, reports GPU memory overhead, and ablate the classification loss weight and BERT unfreezing. The writing is honest about what the LLM input is, which is how I could diagnose the confound so quickly. This is not a sloppy paper; it's an under-controlled one.

Who should read it? People working on food segmentation or on injecting LLM-derived side information into vision models. It's a good starting point for a workshop paper or a short conference paper—after the control experiment is run. The methodological idea, avoiding paired data via LLM labels, is worth discussing even if the specific numbers don't survive unchanged.

My recommendation: yes, send it to peer review, because the idea is interesting and the modules are plausible. But reviewers should push hard for a label-conditioned control. Without it, the central claim remains underdetermined, and I would not accept the paper in its current form.

Referee Report

3 major / 4 minor

Summary. The paper proposes two plug-and-play language-injection modules, LIM-F and LIM-Q, for food image segmentation. Ingredient labels are generated from the input image by an LLM (GPT-o4 mini), encoded by BERT, and injected either into multi-scale visual features (LIM-F) or into Mask2Former decoder queries (LIM-Q). On FoodSeg103 with a Swin-L backbone, LIM-Q reaches a reported mIoU of 55.0, LIM-F reaches 54.4, and the paper also reports gains on rare classes and moderate training-memory overhead.

Significance. If the module-level attribution were established, this would be a practical contribution: it avoids paired image-text pretraining, is architecture-agnostic, and reports modest memory costs. Strengths include comparisons with several strong baselines, ablations of classification-loss weights and BERT fine-tuning, and qualitative examples. However, the evaluation currently cannot separate the effect of the LIM cross-attention design from the effect of the privileged, image-specific ingredient-presence signal that only the LIM variants receive. Because a trivial label-conditioning baseline is missing, the central claim that the LIM modules themselves cause the observed mIoU gains is underdetermined.

major comments (3)

[§4.1, Table 2] The LLM input is "a food image and the full list of 104 ingredient class names" (Section 4.1), so every LIM variant receives an image-specific semantic prior (e.g., "date is present") that is not provided to the Mask2Former, K-Net, BEiT, or Swin-TUNA baselines. The reported rare-class jumps, especially date from 1.28 to 88.93 in Table 2, are exactly the pattern expected from a class-presence prior rather than from the specific LIM cross-attention modules. The paper needs a control that injects the same LLM labels into the baseline through a trivial mechanism (e.g., per-image class-prior logit adjustment, a label-conditioned classifier head, or a query-bias initialization) and reports mIoU and per-class results. Without such a control, the gains cannot be attributed to LIM-F/LIM-Q.
[Abstract vs. Table 1] The abstract claims an improvement from 47.7 to 49.8 mIoU for the CNN-based K-Net architecture, but Table 1 reports 49.0 (+2.7%) for K-Net + LIM-F. This internal inconsistency needs correction. In addition, no error bars, multiple seeds, or significance tests are reported; the per-class numbers in Table 2 are single runs with very large swings. The quantitative claims should be made statistically reliable.
[§4.1, Reproducibility] LLM-generated labels are a central input to the method, but the description is insufficient for reproduction: the model name "GPT-o4 mini", access date, exact prompt template, sampling parameters, and the actual generated-label outputs are not provided, and no code is released. At minimum, the authors should provide the full prompt, temperature/token settings, a sample of LLM outputs, and a cached label file, so the language-guidance input can be verified and reused.

minor comments (4)

[§3.1, §4.1] The model name "GPT-o4 mini" appears twice; if GPT-4o mini is intended, please correct the typo. The footnote "accessed in early May, 2025" should include the exact model version and snapshot.
[Table 3] Table 3 reports only a subset of loss-weight settings. For completeness, please include the setting (5,5,2) with BERT unfreezing for LIM-F and LIM-Q, so the effect of BERT unfreezing is not aliased with the loss-weight change.
[Figure 4] The caption reports final mIoU values 54.44 and 54.96, which are consistent with Table 1 but would be clearer if they matched exactly; round consistently to one decimal place.
[§2.3] The related-work discussion lists LLM vision systems (MiniGPT-4, LLaVA, KOSMOS-2) without clarifying how they differ from the proposed use of an LLM as a label generator; a sentence distinguishing label generation from end-to-end multimodal pretraining would help.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: LLM-derived label input creates an evaluation confound, but no derivation step reduces to its inputs by construction.

full rationale

This paper makes an empirical claim (the LIM modules improve mIoU on FoodSeg103) rather than a first-principles derivation. The proposed LIM-F and LIM-Q are defined by concrete cross-attention operations with BERT-encoded ingredient embeddings; their reported gains are experimental outcomes, not quantities constructed from the inputs. I checked the specified circularity patterns. There are no load-bearing self-citations: references to Mask2Former, Swin, FoodSeg103, OVFoodSeg, and ReLeM are external prior work, and no uniqueness theorem or prior result by these authors is invoked to force the design. The LLM label generation in Section 4.1 does give LIM variants an extra, image-specific class-presence signal that the Mask2Former, K-Net, BEiT, and Swin-TUNA baselines do not receive, so the mIoU difference does not isolate the LIM module's contribution; a label-conditioned control (e.g., injecting the same LLM labels through a trivial query bias or class-prior logit adjustment) is missing. That is a validity/control problem, but it is not a case where a 'prediction' is equivalent by construction to an input: the LLM labels are not the ground-truth masks, the reported mIoU is computed against FoodSeg103 ground truth, and nothing in the paper defines the modules in terms of the mIoU. The lambda_cls and BERT-unfreezing choices in Table 3 are test-set hyperparameter selections, which is a reporting/overfitting concern, not self-definitional circularity. The abstract/table discrepancy for the CNN variant (49.8 vs 49.0) is a typo-level inconsistency, not a circular step. No equation in the paper reduces to its own input. Therefore score 0.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The method rests on two design bets: (1) that LLM-generated ingredient names are a useful and available signal at inference, and (2) that the proposed cross-attention modules, not the label information itself, drive the gains. The second is untested because no label-only control is run. Loss weights and BERT unfreezing are tuned on the validation set. No new physical entities are introduced.

free parameters (2)

classification loss weight lambda_cls = 4 (LIM-F), 6 (LIM-Q)
Table 3: best results require increasing lambda_cls from the Mask2Former default of 2; values are selected on the validation benchmark.
BERT last-4-layer unfreezing = True for best results
Table 3: unfreezing BERT's last four layers raises mIoU from 52.1 to 54.4 (LIM-F) and from 52.8 to 55.0 (LIM-Q); this choice is tuned post hoc.

assumptions (4)

domain assumption GPT-o4 mini, prompted with the input image and the 104 FoodSeg103 class names, returns useful ingredient lists at train and test time.
Section 4.1: the LLM labels are the language signal for both modules; if the LLM fails on a class, the module has no information to guide it.
domain assumption BERT-large-uncased embeddings of single ingredient class names retain enough semantics to guide pixel-level segmentation.
Sections 3.2 and 3.3: BERT vectors are used as keys and values in cross-attention; no evidence is given that class-name embeddings encode visual distinctions like tofu vs cheese.
ad hoc to paper Any improvement over baselines is caused by the LIM cross-attention modules, not merely by the extra class-presence information.
The missing control baseline makes this an unverified premise; Table 2's large rare-class gains are consistent with a label-prior effect rather than the module design.
domain assumption FoodSeg103 mIoU with the standard split is the right yardstick for claiming state-of-the-art.
Section 4.2, Table 1: all comparisons use this benchmark; no other dataset, cross-domain, or open-vocabulary evaluation is attempted.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Food Image Segmentation with LLM-Derived Ingredient Labels and Multimodal Fusion." pith.science (2026). https://pith.science/paper/HMQUKZ5X

@misc{pith2026260725820,
  author       = {Pith},
  title        = {Pith review of: Food Image Segmentation with LLM-Derived Ingredient Labels and Multimodal Fusion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HMQUKZ5X}},
  note         = {Machine review of arXiv:2607.25820}
}

read the original abstract

Food image segmentation plays a vital role in health-related applications such as nutrition tracking and personalized health monitoring. However, existing models often underperform on visually similar ingredients and rare food categories. To address this issue, we propose two plug-and-play multimodal modules that enhance the segmentation performance by leveraging ingredient labels inferred from food images using large language models (LLMs). The first module, called LIM-F (Language Injection Module for Features), is designed to pair with any image encoder that produces multi-layer outputs (e.g., Swin Transformer), while the second module, LIM-Q (Language Injection Module for Queries), targets Mask2Former-style Transformer-based decoders. Both modules enable training without the need for pre-aligning images with text by directly injecting semantic ingredient information into the visual analysis pipeline. On the FoodSeg103 benchmark, the proposed method achieves state-of-the-art performance. Specifically, integrating LIM-Q into the Mask2Former decoder with a Swin-L image encoder yields a mean Intersection over Union (mIoU) of 55.0. LIM-F also demonstrates strong generalization and competitive performance, reaching an mIoU of 54.4 under the same model (Swin-L+Mask2Former). Furthermore, its applicability extends beyond Transformer-based decoders, as evidenced by an improvement from 47.7 to 49.8 mIoU when integrated into a CNN-based architecture. Notably, the improved segmentation accuracy is achieved with only a moderate (at most 3.8 GB) increase in the GPU memory consumption during training. Thus, the proposed approach offers a practical and scalable solution for fine-grained food understanding.

Figures

Figures reproduced from arXiv: 2607.25820 by the authors.

**Figure 1.** Overview of the proposed multimodal segmentation framework. An input image is sent to both a vision encoder and an LLM, which generates ingredient labels that are then encoded by BERT. The language features are injected into the segmentation pipeline via either LIM-F or LIM-Q. LIM-F fuses feature maps after the encoder, while LIM-Q embeds language information into the decoder. Both the encoder and decoder are modula… view at source ↗

**Figure 2.** The segmentation framework with the LIM-F module. Feature maps from different levels are individually passed through a dedicated LIM-F block. These blocks apply cross-attention between the image features and BERT-encoded ingredient embeddings, allowing multi-scale semantic fusion before passing to the decoder. All query, key, and value vectors are layer-normalized to ensure stable feature distributions. the decode… view at source ↗

**Figure 3.** Each LIM-Q block is inserted into every decoding round of Mask2Former to enhance its ability to incorporate textual semantics. It includes a cross-attention to BERT-encoded ingredient embeddings, followed by self-attention and a feed-forward network (FFN). All Q, K, and V inputs are layer-normalized before entering each attention block. throughout decoding, effectively reminding the queries of the ingredient catego… view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: This figure compares the evolution of mIoUs over the training process, with the baseline Mask2Former, and its variants with LIM-F and LIM-Q modules. The two LIM-augmented models consistently outperform the baseline across training iterations, achieving higher final mIo…

**Figure 5.** Figure 5: Sample segmentation results. 5 Conclusion This study introduces two lightweight, plug-and-play multimodal fusion modules, LIM-F and LIM-Q, that enhance food image segmentation through language guidance. LIM-F integrates linguistic semantics into the intermediate vis…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

25 extracted references · 8 linked inside Pith

[1]

In: arXiv:2507.17347 (2025)

Chen, H., Xiao, Z.: Swin-tuna: A novel peft approach for accurate food image segmentation. In: arXiv:2507.17347 (2025)

arXiv 2025
[2]

Journal of Imaging10(12) (2024)

Chen, L., Wang, B., Zhang, J.: Ingredsam: Open-world food ingredient segmenta- tion via a single image prompt. Journal of Imaging10(12) (2024)

2024
[3]

IEEE Transactions on Pattern Analysis and Machine Intelli- gence40(4), 834–848 (2017)

Chen, L.C., Papandreou, G., Kokkinos, I., Murphy, K., Yuille, A.L.: Deeplab: Se- mantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs. IEEE Transactions on Pattern Analysis and Machine Intelli- gence40(4), 834–848 (2017)

2017
[4]

In: Proceedings of European Conference on Computer Vision (2020)

Chen, Y.C., Li, L., Yu, L., Kholy, A.E., Ahmed, F., Gan, Z., Cheng, Y., Liu, J.: Uniter: Universal image-text representation learning. In: Proceedings of European Conference on Computer Vision (2020)

2020
[5]

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

Cheng, B., Misra, I., Schwing, A.G., Kirillov, A., Girdhar, R.: Masked-attention mask transformer for universal image segmentation. In: Proceedings of IEEE/CVF International Conference on Computer Vision and Pattern Recognition. pp. 1290– 1299 (2022)

2022
[6]

Chiang, W.L., Li, Z., Lin, Z., Sheng, Y., Wu, Z., Zhang, H., Zheng, L., Zhuang, S., Zhuang, Y., Gonzalez, J.E., Stoica, I., Xing, E.P.: Vicuna: An open-source chatbot impressing gpt-4 with 90%* chatgpt quality (2023), https://lmsys.org/blog/2023- 03-30-vicuna/

2023
[7]

In: Proceedings of Annual Con- ference of the North American Chapter of the Association for Computational Lin- guistics: Human Language Technologies

Devlin, J., Chang, M.W., Lee, K., Toutanova, K.: Bert: Pre-training of deep bidi- rectional transformers for language understanding. In: Proceedings of Annual Con- ference of the North American Chapter of the Association for Computational Lin- guistics: Human Language Technologies. pp. 4171–4186 (2019)

2019
[8]

In: Proceedings of International Conference on Learning Representations (2021)

Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., Uszkoreit, J., Houlsby, N.: An image is worth 16x16 words: Transformers for image recognition at scale. In: Proceedings of International Conference on Learning Representations (2021)

2021

Show all 25 references

[9]

In: arXiv:2304.02643 (2023) Title Suppressed Due to Excessive Length 13

Kirillov, A., Mintun, E., Ravi, N., Mao, H., Rolland, C., Gustafson, L., Xiao, T., Whitehead, S., Berg, A.C., Lo, W.Y., Dollar, P., Girshick, R.: Segment anything. In: arXiv:2304.02643 (2023) Title Suppressed Due to Excessive Length 13

2023 arXiv
[10]

In: arXiv:2308.05938 (2023)

Lan, X., Lyu, J., Jiang, H., Dong, K., Niu, Z., Zhang, Y., Xue, J.: Foodsam: Any food segmentation. In: arXiv:2308.05938 (2023)

2023 arXiv
[11]

In: Proceedings of Advances in Neural Information Processing Systems (2023)

Liu, H., Li, C., Wu, Q., Lee, Y.J.: Visual instruction tuning. In: Proceedings of Advances in Neural Information Processing Systems (2023)

2023
[12]

Liu, Z., Lin, Y., Cao, Y., Hu, H., Wei, Y., Zhang, Z., Lin, S., Guo, B.: Swin transformer:Hierarchical visiontransformerusingshiftedwindows.In:Proceedings of IEEE/CVF International Conference on Computer Vision (2021)

2021
[13]

In: Proceedings of International Conference on Learning Representations (2017)

Loshchilov, I., Hutter, F.: Sgdr: Stochastic gradient descent with warm restarts. In: Proceedings of International Conference on Learning Representations (2017)

2017
[14]

In: Proceedings of International Conference on Neural Information Processing Systems (2019)

Lu,J.,Batra,D.,Parikh,D.,Lee,S.:Vilbert:Pretrainingtask-agnosticvisiolinguis- tic representations for vision-and-language tasks. In: Proceedings of International Conference on Neural Information Processing Systems (2019)

2019
[15]

IEEE Transactions on Pattern Analysis and Machine Intelligence43(1), 187–203 (2021)

Marin, J., Biswas, A., Ofli, F., Hynes, N., Salvador, A., Aytar, Y., Weber, I., Tor- ralba, A.: Recipe1m+: A dataset for learning cross-modal embeddings for cooking recipes and food images. IEEE Transactions on Pattern Analysis and Machine Intelligence43(1), 187–203 (2021)

2021
[16]

In: arXiv:2303.08774 (2023)

OpenAI: Gpt-4 technical report. In: arXiv:2303.08774 (2023)

2023 arXiv
[17]

In: arXiv:2306.14824 (2023)

Peng, Z., Wang, W., Dong, L., Hao, Y., Huang, S., Ma, S., Wei, F.: Kosmos-2: Grounding multimodal large language models to the world. In: arXiv:2306.14824 (2023)

2023 arXiv
[18]

In: arXiv:2103.00020 (2021)

Radford, A., Kim, J.W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., Krueger, G., Sutskever, I.: Learning transferable visual models from natural language supervision. In: arXiv:2103.00020 (2021)

2021 arXiv
[19]

In: arXiv:2306.09203 (2023)

Sinha, G., Parmar, K., Azimi, H., Tai, A., Chen, Y., Wong, A., Xi, P.: Transferring knowledge for food image segmentation using transformers and convolutions. In: arXiv:2306.09203 (2023)

2023 arXiv
[20]

In: Proceedings of ACM International Conference on Multimedia

Wu, X., Fu, X., Liu, Y., Lim, E.P., Hoi, S.C., Sun, Q.: A large-scale benchmark for food image segmentation. In: Proceedings of ACM International Conference on Multimedia. pp. 506–515 (2021)

2021
[21]

In: Proceedings of IEEE/CVF International Conference on Computer Vision and Pattern Recogni- tion

Wu, X., Yu, S., Lim, E.P., Ngo, C.W.: Ovfoodseg: Elevating open-vocabulary food image segmentation via image-informed textual representation. In: Proceedings of IEEE/CVF International Conference on Computer Vision and Pattern Recogni- tion. pp. 4144–4153 (2024)

2024
[22]

In: Proceedings of European Conference on Computer Vision (2018)

Xiao, T., Liu, Y., Zhou, B., Jiang, Y., Sun, J.: Unified perceptual parsing for scene understanding. In: Proceedings of European Conference on Computer Vision (2018)

2018
[23]

In: Proceedings of Annual Conference on Neural Information Processing Systems (2021)

Zhang, W., Pang, J., Chen, K., Loy, C.C.: K-net: Towards unified image segmen- tation. In: Proceedings of Annual Conference on Neural Information Processing Systems (2021)

2021
[24]

In: Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (2017)

Zhao, H., Shi, J., Qi, X., Wang, X., Jia, J.: Pyramid scene parsing network. In: Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (2017)

2017
[25]

In: arXiv:2304.10592 (2023)

Zhu, D., Chen, J., Shen, X., Li, X., Elhoseiny, M.: Minigpt-4: Enhanc- ing vision-language understanding with advanced large language models. In: arXiv:2304.10592 (2023)

2023 arXiv

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.

[1] [1]

In: arXiv:2507.17347 (2025)

Chen, H., Xiao, Z.: Swin-tuna: A novel peft approach for accurate food image segmentation. In: arXiv:2507.17347 (2025)

arXiv 2025

[2] [2]

Journal of Imaging10(12) (2024)

Chen, L., Wang, B., Zhang, J.: Ingredsam: Open-world food ingredient segmenta- tion via a single image prompt. Journal of Imaging10(12) (2024)

2024

[3] [3]

IEEE Transactions on Pattern Analysis and Machine Intelli- gence40(4), 834–848 (2017)

Chen, L.C., Papandreou, G., Kokkinos, I., Murphy, K., Yuille, A.L.: Deeplab: Se- mantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs. IEEE Transactions on Pattern Analysis and Machine Intelli- gence40(4), 834–848 (2017)

2017

[4] [4]

In: Proceedings of European Conference on Computer Vision (2020)

Chen, Y.C., Li, L., Yu, L., Kholy, A.E., Ahmed, F., Gan, Z., Cheng, Y., Liu, J.: Uniter: Universal image-text representation learning. In: Proceedings of European Conference on Computer Vision (2020)

2020

[5] [5]

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

Cheng, B., Misra, I., Schwing, A.G., Kirillov, A., Girdhar, R.: Masked-attention mask transformer for universal image segmentation. In: Proceedings of IEEE/CVF International Conference on Computer Vision and Pattern Recognition. pp. 1290– 1299 (2022)

2022

[6] [6]

Chiang, W.L., Li, Z., Lin, Z., Sheng, Y., Wu, Z., Zhang, H., Zheng, L., Zhuang, S., Zhuang, Y., Gonzalez, J.E., Stoica, I., Xing, E.P.: Vicuna: An open-source chatbot impressing gpt-4 with 90%* chatgpt quality (2023), https://lmsys.org/blog/2023- 03-30-vicuna/

2023

[7] [7]

In: Proceedings of Annual Con- ference of the North American Chapter of the Association for Computational Lin- guistics: Human Language Technologies

Devlin, J., Chang, M.W., Lee, K., Toutanova, K.: Bert: Pre-training of deep bidi- rectional transformers for language understanding. In: Proceedings of Annual Con- ference of the North American Chapter of the Association for Computational Lin- guistics: Human Language Technologies. pp. 4171–4186 (2019)

2019

[8] [8]

In: Proceedings of International Conference on Learning Representations (2021)

Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., Uszkoreit, J., Houlsby, N.: An image is worth 16x16 words: Transformers for image recognition at scale. In: Proceedings of International Conference on Learning Representations (2021)

2021