Pith. sign in

REVIEW 4 major objections 5 minor 67 references

Human-Guided Image Generation for Expanding Small-Scale Training Image Datasets

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read Human-guided dataset expansion, in which users explore generated and original images together and give feedback on individual samples, lifts classification accuracy from 48.45% to 81.80% and detection mAP from 92.4% to 94.5%, the paper…

desk verdict M2M projection is a real contribution; the prompt-refinement proxy is the weak link and the paper would benefit from a controlled validation. read the letter →

arxiv 2412.16839 v2 pith:VB6L2LWS submitted 2024-12-22 cs.CV cs.AI

classification cs.CVcs.AI
keywords datasetexpansionhuman-guidedimagegenerationmulti-modalprojectioncontrastivelearningpromptrefinementvisualanalyticssmall-scaledatasetsStableDiffusion
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's central claim is that the limiting factor in generative dataset expansion is control, not generation: automatically produced images are insufficiently diverse and contain undesired content, and users cannot easily repair them by editing prompts. To make expansion controllable, the paper builds DataCrafter, a visual analytics workflow in which a contrastive multi-modal projection places images and their content labels in a shared 2D view, and sample-level feedback--deleting undesired images or selecting images for more variety--drives an evolutionary prompt refiner that recommends better prompts. In a ten-class pet-image case study, the workflow raises a classifier's accuracy from 48.45% to 81.80%, about five percentage points above an automatic expansion baseline, and raises object-detection mAP from 92.4% to 94.5% when the expanded set is used to retrain a detector. A sympathetic reader should care because small labelled datasets are a practical bottleneck, and the paper offers a concrete way to put a human in the loop without requiring prompt-engineering skill.

What carries the argument

Two coupled mechanisms carry the argument. The first is M2M, a six-layer feed-forward projection network trained to minimize a weighted sum of three contrastive losses--image-to-image, image-to-label, and label-to-label--with frequency-biased negative sampling that samples absent labels by their frequency to prevent rare labels from collapsing. It produces the shared scatterplot in which images and content labels are explored. The second is sample-level prompt refinement, an evolutionary algorithm that asks an LLM to mutate a prompt, generates a few proxy images with the mutated prompt, scores them against the user's feedback (away from deleted images and close to kept ones, or diverse and close to selected ones, with high confidence), and iterates until the score stops improving. The projection supplies the evidence a user needs to give feedback; the refinement supplies the mechanism that turns feedback into new prompts.

What would settle it

Generate two batches of images for the same class with equal proxy scores, one from DataCrafter's refined prompts and one from arbitrary prompts that happen to match those scores; if the high-scoring batch does not train a better downstream classifier than the low-scoring one, the refinement objective is not the cause of the reported gains. A simpler check: take the proxy images a refined prompt scores highest on and train a model on them alone; if accuracy falls below training on images the proxy scores lowest on, the load-bearing premise fails.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that many-to-many multi-modal projection--the common situation where one image carries several content labels--cannot in general preserve exact distance orders between images and labels. The paper proves the number of achievable orders is at most $n(n-1)(n^2-n+2)/8+1$, which for $n>3$ is less than $n!$; hence no single planar mapping can satisfy the distance-order loss used by the existing MFM method, and that loss fails exactly in the setting that matters for dataset exploration. The replacement is M2M, a contrastive projection that keeps similar images together, similar labels together, and images near their labels. Embedded in DataCrafter, this projection lets a user spot that the class name 'Bengal' is generating cat-tiger hybrids, remove those images with one click, and accept a refined prompt that fixes the class name; combined with a diversity-increasing refinement step, the final expanded dataset lifts classification accuracy from 48.45% to 81.80% and improves detection mAP from 92.4% to 94.5%.

Load-bearing premise

The loop assumes that a prompt scoring well on the proxy objectives--similarity to desired images, diversity, and confidence of its generated proxy images--will actually produce images that improve the downstream model; the paper does not validate that correlation.

Editorial extensions

If this is right

  • Small labelled datasets can be expanded with generative models in a controlled way, and the resulting classifiers beat fully automatic expansion on the same budget.
  • The many-to-many projection result implies that any future multi-modal projection method for multi-label image collections must handle the impossibility of exact distance-order preservation, and contrastive objectives are a viable replacement.
  • Users who cannot write good prompts can still steer generation by pointing at images, because the sample-level refinement loop converts deletions and selections into prompt changes.
  • The same expanded dataset transfers to a second downstream task (object detection), so the benefit is not specific to one classifier architecture.

Reading between the lines

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

  • The paper leaves untested whether the proxy objectives in the refinement loop (similarity, diversity, confidence on a few proxy images) are what cause the downstream accuracy gain; a controlled study that optimizes the same objectives without human feedback would isolate the mechanism.
  • The frequency-biased negative sampling idea is a generally useful trick for contrastive learning on long-tailed label distributions, beyond the projection setting in which it appears.
  • The paper's own limitation section (Sec. 7.2) notes that feedback is currently single-set, that weaker generators for other image types may hurt quality, and that user efficiency was not measured; these caveats bound, but do not undo, the reported gains.
  • A natural extension the paper mentions but does not develop is feedback across multiple sets of generated images, such as fusing styles of two sets; this could further increase diversity.
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

4 major / 5 minor

Summary. The paper proposes DataCrafter, an interactive visual analytics system for expanding small image datasets with a text-to-image generative model under human guidance. It introduces two technical components: M2M, a contrastive-learning-based multi-modal projection method for jointly embedding images and content labels in 2D, and a sample-level prompt refinement method that uses an evolutionary algorithm to turn user feedback (deleting undesired images or requesting more images like selected ones) into revised prompts. The evaluation consists of a quantitative comparison of M2M against single-modal and multi-modal projection baselines on Pets and COCO, a single-expert case study on a 10-class Pets subset showing classification accuracy improving from 48.45% to 81.80%, a comparison with the automatic GIF expansion method, and an extension to object detection. The paper claims that M2M preserves intra- and inter-modal relationships better than existing projection methods and that sample-level feedback makes dataset expansion more controllable and effective.

Significance. If the claims are substantiated, DataCrafter is a useful contribution to interactive dataset generation: it addresses the real problem of expanding small training sets and lowers the barrier to prompt refinement by letting users give image-level feedback. The M2M formulation, especially the frequency-biased negative sampling for contrastive loss, is a sensible adaptation of contrastive learning to multi-modal projection, and the authors provide source code and data, which aids reproducibility. The reported case-study gain (33.35 percentage points) is large and the qualitative finding that the class name 'Bengal' invited tiger-like generations is compelling. However, the significance is currently conditional: the theoretical analysis does not actually guarantee the superiority of the contrastive objective, the projection evaluation lacks statistical reliability, and the benefit of the prompt-refinement algorithm is not isolated from the human expert's inspection and decisions.

major comments (4)
  1. [Sec. 5.3.1, Eqs. (8)-(9)] The sample-level prompt refinement selects mutated prompts by maximizing proxy objectives defined in CLIP space (similarity to remaining/selected images, diversity, and zero-shot confidence) evaluated on a small number of stochastic proxy images. No experiment validates that optimizing these proxies improves the actual downstream objective, namely the accuracy of a ResNet50 trained on the expanded dataset. The case study in Sec. 6.2 confounds the proxy with E1's own inspection and prompt acceptance/rejection decisions, so the 81.80% accuracy cannot be attributed to the objective in Eqs. (8)-(9). Without an ablation that isolates the prompt-refinement optimizer (e.g., comparing prompts recommended by the objective against random mutations for the same feedback, or measuring correlation between the proxy score and downstream accuracy), the central claim that sample-level feedback yields better prompts rests on a single anecdote.
  2. [Sec. 5.2.1, Theorems 1-2 and Lemma 3] The theoretical analysis shows only that MFM's exact multi-modal distance-order loss cannot always be satisfied in the many-to-many setting when n > 3. It does not provide any guarantee, approximation bound, or theoretical characterization for the proposed contrastive objective in Eqs. (5)-(6). The sentence in Sec. 6.1, 'This validates the correctness of our theoretical analysis and demonstrates the effectiveness of contrastive learning in multi-modal projection,' overstates the logical link, since the theorems neither predict nor explain the empirical success of M2M. The contribution statement that the projection method is 'theoretically and experimentally better than the existing methods' is therefore too strong; the authors should either add a theorem or explicit theoretical justification for contrastive loss in this setting or revise the claim to empirical superiority only.
  3. [Sec. 6.1, Table 1] All quantitative results in Table 1 are single point estimates without error bars, number of runs, or significance tests. Because M2M is trained with stochastic mini-batch sampling and frequency-biased negative sampling, the observed margins over MFM (e.g., IMS 0.7311 vs. 0.5052 on Pets) could vary across random seeds. The paper claims that M2M 'consistently outperforms' baselines, but with one run per method the consistency cannot be verified. The authors should report means and standard deviations over at least three random seeds and, where feasible, a paired significance test for the difference against MFM.
  4. [Sec. 6.3, Table 2] The comparison with the automatic expansion method GIF is not a controlled test of the human-guided process. The row 'Dataset expansion with human guidance' uses prompts that were modified by E1 during the interactive case study, so the 5% improvement over GIF reflects the quality of the human-curated prompts, not the ability of DataCrafter's sample-level prompt refinement to produce them automatically. Additionally, the case study involved only one expert (E1), who also participated in the requirement analysis, so the improvement may not generalize to other users. An evaluation with multiple users, or at least a separate test in which the prompt-refinement algorithm receives the same sample-level feedback without human inspection of intermediate results, is needed to support the claim that the tool itself provides an advantage over automatic expansion.
minor comments (5)
  1. [Sec. 4] The sentence 'Both E3 and R4 are actively involved in deploying data expansion methods' contains a typo: 'R4' should be 'E4'.
  2. [Sec. 5.2.1] The phrase 'we develop a conservative learning-based multi-modal projection method' appears to be a typo for 'contrastive learning-based'; please correct it.
  3. [Sec. 5.2.1, Eq. (4)] In Eq. (4), the distance functions h(·, ·) and l(·, ·) are described only in the surrounding prose; they should be defined formally in the equation or immediately before it for readability.
  4. [Sec. 5.1, Eq. (3)] The CMMD distance in Eq. (3) includes a Gaussian kernel scale σ, but the choice of σ is never stated; please specify the value or the heuristic used for setting it.
  5. [Sec. 6.3] In the object detection extension, the pseudo-labels are generated by the detector trained on the original images and then used as ground truth for retraining; this self-training protocol may yield optimistic mAP estimates. Please report the quality of the pseudo-labels (e.g., precision on a labeled subset) or use a held-out validation set to confirm the improvement.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: downstream accuracy is measured on a held-out test set, and no fitted value of the central claim is recycled as evidence.

full rationale

DataCrafter's central empirical claim is that human-guided expansion improves downstream classifiers (Table 2: 48.45% to 81.80% on the Pets test set, about 5% above the GIF baseline). This accuracy is measured on a held-out test set and is not an input to any fitted parameter of the M2M projection or the prompt-refinement method. The M2M projection is an unsupervised dimensionality-reduction method trained on CLIP embeddings of images and extracted content labels; it is compared against baselines (t-SNE, UMAP, DCM, MFM) using trustworthiness, continuity, and the paper's IMS measure. Although IMS is aligned with M2M's image-label contrastive objective, this is standard practice for evaluating unsupervised embeddings and does not make the downstream accuracy claim circular. The sample-level prompt refinement (Eqs. 8-9) optimizes proxy objectives on a few generated proxy images; the paper does not demonstrate that these proxies correlate with downstream accuracy, which is a validity/robustness limitation, not circularity, because accuracy is measured independently on the test set and the expert's accept/reject decisions are not fitted to that test set. Theoretical claims about order preservation (Theorem 2 and Lemma 3) are argued from combinatorial counting rather than assumed from self-citation. No load-bearing self-citation chain was found: the cited GIF method [65] is used as a baseline and metric source, but the comparison is external and the central result does not reduce to that citation.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The central result relies on several unverified practical assumptions: CLIP and GPT-4 must provide semantically meaningful labels, Stable Diffusion must respond to prompts as expected, and proxy-image objectives must predict downstream accuracy. These are not new physical entities, but they are load-bearing domain assumptions. The listed hyperparameters are free values that the paper does not report, making exact reproduction difficult.

free parameters (6)
  • Contrastive temperature tau
    Appears in Eq. 6; no value is reported, and M2M projection quality depends on it.
  • Nearest-neighbor count k
    Defines positive pairs for image-image and label-label contrastive losses in Sec. 5.2.1; value not reported.
  • Loss weights omega_1 and omega_2
    Balance the three contrastive losses in Eq. 5; the paper says they are set by multi-task learning [37] but gives no values.
  • CMMD kernel scale sigma
    Gaussian kernel in Eq. 3 used for the distance metric; no value is given.
  • Number of proxy images N_g
    Number of images generated to evaluate each mutated prompt in Sec. 5.3.1; not specified.
  • Confidence function and threshold
    Confidence(I_g^i) appears in Eqs. 8 and 9 but the paper does not define which classifier produces it or how it is calibrated.
assumptions (6)
  • domain assumption CLIP embeddings place images and text labels in a shared space where distance reflects semantic similarity.
    Used throughout Sec. 5.1 and 5.2 for content labels, metrics, and M2M; if this fails, projection and evaluation lose meaning.
  • domain assumption GPT-4 captions and NLTK noun extraction provide accurate content labels.
    Sec. 6.1 extracts labels this way; all exploration and prompt decisions in the case study depend on label quality.
  • domain assumption Stable Diffusion can be steered by prompt templates to generate task-relevant, diverse images from few seed images.
    Fundamental premise of dataset expansion in Sec. 3 and Sec. 5.3.
  • domain assumption A few proxy images generated under a mutated prompt faithfully indicate the quality of future images from that prompt.
    The evolutionary refinement in Sec. 5.3.1 accepts or rejects mutations based only on proxy images.
  • domain assumption High zero-shot confidence is a reliable proxy for usefulness in downstream training (large margin theory).
    Confidence appears as an additive reward in Eqs. 8 and 9; no experiment validates this connection to downstream accuracy.
  • domain assumption Loss-balancing multi-task learning [37] correctly determines omega_1 and omega_2 in Eq. 5.
    The authors state the weights are determined by [37] but provide no values or verification that the resulting balance is appropriate.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Human-Guided Image Generation for Expanding Small-Scale Training Image Datasets." pith.science (2026). https://pith.science/paper/VB6L2LWS

@misc{pith2026241216839,
  author       = {Pith},
  title        = {Pith review of: Human-Guided Image Generation for Expanding Small-Scale Training Image Datasets},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VB6L2LWS}},
  note         = {Machine review of arXiv:2412.16839}
}
read the original abstract

The performance of computer vision models in certain real-world applications (e.g., rare wildlife observation) is limited by the small number of available images. Expanding datasets using pre-trained generative models is an effective way to address this limitation. However, since the automatic generation process is uncontrollable, the generated images are usually limited in diversity, and some of them are undesired. In this paper, we propose a human-guided image generation method for more controllable dataset expansion. We develop a multi-modal projection method with theoretical guarantees to facilitate the exploration of both the original and generated images. Based on the exploration, users refine the prompts and re-generate images for better performance. Since directly refining the prompts is challenging for novice users, we develop a sample-level prompt refinement method to make it easier. With this method, users only need to provide sample-level feedback (e.g., which samples are undesired) to obtain better prompts. The effectiveness of our method is demonstrated through the quantitative evaluation of the multi-modal projection method, improved model performance in the case study for both classification and object detection tasks, and positive feedback from the experts.

Figures

Figures reproduced from arXiv: 2412.16839 by the authors.

Figure 1
Figure 1. The comparison between the automatic and the proposed human-guided dataset expansion methods. (a) The automatic dataset [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Dataset expansion consists of two main steps: (a) latent pertur [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Method overview: (a) given a small number of original images and prompts, a set of images is generated; (b)-(d) three visualizations to [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: DataCrafter: (a) an information panel to show the numbers of original and generated images and content labels; (b) a metric visualization [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: The evolutionary-algorithm-based prompt refinement. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Visual comparison between our method and three existing methods. [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: The prompt templates and multi-modal distribution visualization (left) before improving diversity and (right) after improving diversity. [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

67 extracted references · 31 canonical work pages

  1. [1]

    Arias-Hernandez, L

    R. Arias-Hernandez, L. T. Kaastra, T. M. Green, and B. Fisher. Pair analytics: Capturing reasoning processes in collaborative visual analytics. In IEEE Hawaii International Conference on System Sciences, pp. 1–10,

  2. [2]

    Borg and P

    I. Borg and P. J. Groenen. Modern multidimensional scaling: Theory and applications. Springer Science & Business Media, 2007. doi: 10.1111/j. 1745-3984.2003.tb01108.x 7

  3. [3]

    B. E. Boser, I. M. Guyon, and V . N. Vapnik. A training algorithm for opti- mal margin classifiers. In Proceedings of the Workshop on Computational Learning Theory, pp. 144–152, 1992. doi: 10.1145/130385.130401 6

  4. [4]

    Brade, B

    S. Brade, B. Wang, M. Sousa, S. Oore, and T. Grossman. Promptify: Text- to-image generation through interactive prompt exploration with large language models. In Proceedings of the Annual ACM Symposium on User Interface Software and Technology, pp. 1–14, 2023. doi: 10.1145/3586183 .3606725 2

  5. [7]

    C. Chen, Z. Wang, J. Wu, X. Wang, L.-Z. Guo, Y .-F. Li, and S. Liu. Interactive graph construction for graph-based semi-supervised learning. IEEE Transactions on Visualization and Computer Graphics, 27(9):3701– 3716, 2021. doi: 10.1109/TVCG.2021.3084694 2, 5, 6

  6. [8]

    C. Chen, J. Wu, X. Wang, S. Xiang, S.-H. Zhang, Q. Tang, and S. Liu. Towards better caption supervision for object detection.IEEE Transactions on Visualization and Computer Graphics, 28(4):1941–1954, 2022. doi: 10 .1109/TVCG.2021.3138933 1, 2, 9

  7. [9]

    C. Chen, J. Yuan, Y . Lu, Y . Liu, H. Su, S. Yuan, and S. Liu. OoDAnalyzer: Interactive analysis of out-of-distribution samples. IEEE Transactions on Visualization and Computer Graphics, 27(7):3335–3349, 2021. doi: 10. 1109/TVCG.2020.2973258 3, 5

  8. [10]

    Cheng and K

    S. Cheng and K. Mueller. The data context map: Fusing data and attributes into a unified display. IEEE Transactions on Visualization and Computer Graphics, 22(1):121–130, 2015. doi: 10.1109/tvcg.2015.2467552 7

Show all 67 references
  1. [11]

    J. J. Y . Chung and E. Adar. PromptPaint: Steering text-to-image generation through paint medium-like interactions. In Proceedings of the Annual ACM Symposium on User Interface Software and Technology, pp. 23–32,

  2. [12]

    Dunlap, A

    L. Dunlap, A. Umino, H. Zhang, J. Yang, J. E. Gonzalez, and T. Darrell. Diversify your vision datasets with automatic diffusion-based augmenta- tion. In Proceedings of the Advances in Neural Information Processing Systems, pp. 79024–79034, 2023. 1, 3

  3. [13]

    Endert, C

    A. Endert, C. Han, D. Maiti, L. House, and C. North. Observation-level interaction with statistical models for visual analytics. In Proceedings of the IEEE Conference on Visual Analytics Science and Technology, pp. 121–130, 2011. doi: 10.1109/vast.2011.6102449 6

  4. [14]

    Y . Feng, X. Wang, K. K. Wong, S. Wang, Y . Lu, M. Zhu, B. Wang, and W. Chen. PromptMagician: Interactive prompt engineering for text- to-image creation. IEEE Transactions on Visualization and Computer Graphics, 30(1):295–305, 2024. doi: 10.1109/tvcg.2023.3327168 2, 3, 6

  5. [15]

    G. W. Furnas. Generalized fisheye views. ACM Sigchi Bulletin, 17(4):16– 23, 1986. 6

  6. [16]

    L. Gou, L. Zou, N. Li, M. Hofmann, A. K. Shekar, A. Wendt, and L. Ren. V ATLD: A visual analytics system to assess, understand and improve traffic light detection. IEEE Transactions on Visualization and Computer Graphics, 27(2):261–271, 2021. doi: 10.1109/TVCG.2020.3030350 2

  7. [17]

    Q. Guo, R. Wang, J. Guo, B. Li, K. Song, X. Tan, G. Liu, J. Bian, and Y . Yang. Connecting large language models with evolutionary algorithms yields powerful prompt optimizers. InThe Proceedings of the International Conference on Learning Representations, 2024. 6

  8. [18]

    Y . Guo, H. Shao, C. Liu, K. Xu, and X. Yuan. PrompTHis: Visualizing the process and influence of prompt editing during text-to-image creation. IEEE Transactions on Visualization and Computer Graphics, 2024. to be published. doi: 10.1109/tvcg.2024.3408255 2

  9. [19]

    W. Han, C. Kim, D. Ju, Y . Shim, and S. J. Hwang. Advancing text- driven chest x-ray generation with policy-based reinforcement learning. In Proceedings of the International Conference on Medical Image Computing and Computer-Assisted Intervention, pp. 56–66, 2024. 9

  10. [20]

    Hasani, F

    N. Hasani, F. Farhadi, M. A. Morris, M. Nikpanah, A. Rahmim, Y . Xu, A. Pariser, M. T. Collins, R. M. Summers, E. Jones, et al. Artificial intelli- gence in medical imaging and its impact on the rare disease community: threats, challenges and opportunities. PET Clinics, 17(1):...

  11. [21]

    Hausdorff

    F. Hausdorff. Set Theory. Chelsea, 2007. 6

  12. [22]

    J. He, X. Wang, K. K. Wong, X. Huang, C. Chen, Z. Chen, F. Wang, M. Zhu, and H. Qu. VideoPro: A visual analytics approach for interactive video programming. IEEE Transactions on Visualization and Computer Graphics, 30(1):87–97, 2024. doi: 10.1109/tvcg.2023.3326586 2

  13. [23]

    K. He, X. Zhang, S. Ren, and J. Sun. Deep residual learning for im- age recognition. In IEEE Conference on Computer Vision and Pattern Recognition, pp. 770–778, 2016. doi: 10.1109/CVPR.2016.90 9

  14. [25]

    Heusel, H

    M. Heusel, H. Ramsauer, T. Unterthiner, B. Nessler, and S. Hochreiter. GANs trained by a two time-scale update rule converge to a local nash equilibrium. In Proceedings of the 31st International Conference on Neural Information Processing Systems , p. 6629–6640, 2017. doi: 10....

  15. [26]

    Höferlin, R

    B. Höferlin, R. Netzel, M. Höferlin, D. Weiskopf, and G. Heidemann. Inter-active learning of ad-hoc classifiers for video visual analytics. In Proceedings of the Conference on Visual Analytics Science and Technology, pp. 23–32, 2012. doi: 10.1109/vast.2012.6400492 2

  16. [28]

    Jayasumana, S

    S. Jayasumana, S. Ramalingam, A. Veit, D. Glasner, A. Chakrabarti, and S. Kumar. Rethinking fid: Towards a better evaluation metric for image generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 9307–9315, June 2024. 4

  17. [29]

    S. Jia, Z. Li, N. Chen, and J. Zhang. Towards visual explainable active learning for zero-shot classification. IEEE Transactions on Visualization and Computer Graphics, 28(1):791–801, 2022. doi: 10.1109/TVCG.2021. 3114793 5

  18. [30]

    B. Li, X. Xu, X. Wang, Y . Hou, Y . Feng, F. Wang, X. Zhang, Q. Zhu, and W. Che. Semantic-guided generative image augmentation method with diffusion models for image classification. In Proceedings of the AAAI Conference on Artificial Intelligence, pp. 3018–3027, 2024. doi: 10....

  19. [31]

    S. Li, G. Liu, T. Wei, S. Jia, and J. Zhang. EvoVis: A visual analyt- ics method to understand the labeling iterations in data programming. IEEE Transactions on Visualization and Computer Graphics, 2024. to be published. doi: 10.1109/tvcg.2024.3370654 2

  20. [32]

    Y . Li, J. Wang, P. Aboagye, C.-C. M. Yeh, Y . Zheng, L. Wang, W. Zhang, and K.-L. Ma. Visual analytics for efficient image exploration and user- guided image captioning. IEEE Transactions on Visualization and Com- puter Graphics, 2024. to be published. doi: 10.1109/tvcg.2024....

  21. [33]

    T.-Y . Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Dollár, and C. L. Zitnick. Microsoft coco: Common objects in context. In Proceedings of the European Conference on Computer Vision, pp. 740– 755, 2014. doi: 10.1007/978-3-319-10602-1_48 7

  22. [34]

    J. Liu, W. Li, and Y . Sun. Memory-based jitter: Improving visual recog- nition on long-tailed data with diversity in memory. Proceedings of the AAAI Conference on Artificial Intelligence, 36(2):1720–1728, 2022. doi: 10.1609/aaai.v36i2.20064 4

  23. [35]

    M. Liu, J. Shi, Z. Li, C. Li, J. Zhu, and S. Liu. Towards better analysis of deep convolutional neural networks. IEEE Transactions on Visualization and Computer Graphics, 23(1):91–100, 2017. doi: 10.1109/TVCG.2016. 2598831 9

  24. [36]

    S. Liu, C. Chen, Y . Lu, F. Ouyang, and B. Wang. An interactive method to improve crowdsourced annotations. IEEE Transactions on Visualization and Computer Graphics, 25(1):235–245, 2019. doi: 10.1109/TVCG.2018. 2864843 5

  25. [37]

    S. Liu, Y . Liang, and A. Gitter. Loss-balanced task weighting to reduce negative transfer in multi-task learning. In Proceedings of the AAAI Conference on Artificial Intelligence, pp. 9977–9978, 2019. doi: 10.1609/ aaai.v33i01.33019977 5

  26. [38]

    W. Liu, D. Anguelov, D. Erhan, C. Szegedy, S. Reed, C.-Y . Fu, and A. C. Berg. SSD: Single shot MultiBox detector. InProceedings of the European Conference on Computer Vision, pp. 21–37, 2016. 9

  27. [39]

    Loper and S

    E. Loper and S. Bird. NLTK: The natural language toolkit. arxiv preprint arXiv:cs/0205028, 2002. 7

  28. [40]

    F. Lyu, C. Chen, J. Zhang, X. Feng, and Z. Tang. Visualization for supercomputer system: A survey. Journal of Computer-Aided Design and Computer Graphics, 36(3):321–335, 2024. doi: 10.3724/SP.J.1089.2024. 2023-00791 6

  29. [41]

    McInnes, J

    L. McInnes, J. Healy, and J. Melville. UMap: Uniform manifold ap- proximation and projection for dimension reduction. arXiv preprint arXiv:1802.03426, 2018. 7

  30. [42]

    Micallef, G

    L. Micallef, G. Palmas, A. Oulasvirta, and T. Weinkauf. Towards percep- tual optimization of the visual design of scatterplots. IEEE Transactions on Visualization and Computer Graphics, 23(6):1588–1599, 2017. doi: 10 .1109/tvcg.2017.2674978 5

  31. [43]

    GPT-4 technical report

    OpenAI. GPT-4 technical report. arXiv, preprint arXiv: 2303.08774, 2023. 2

  32. [44]

    Parkhi, A

    O. Parkhi, A. Vedaldi, A. Zisserman, and C. Jawahar. Cats and dogs. In Proceedings of the Conference on Computer Vision and Pattern Recogni- tion, pp. 3498–3505, 2012. 7

  33. [45]

    Radford, J

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark, G. Krueger, and I. Sutskever. Learning transferable visual models from natural language supervision. In Proceedings of the International Conference on Machine Learning...

  34. [46]

    C. K. Reddy and B. Vinzamuri. A survey of partitional and hierarchical clustering algorithms. In Data clustering, pp. 87–110. Chapman and Hall/CRC, 2018. doi: 10.1201/9781315373515-4 6

  35. [47]

    Rombach, A

    R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer. High- resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 10684–10695, 2022. doi: 10.1109/cvpr52688.2022.01042 1, 2

  36. [48]

    Shi and Y .-D

    L. Shi and Y .-D. Shen. Diversifying convex transductive experimental design for active learning. In Proceedings of the International Joint Conference on Artificial Intelligence, pp. 1997–2003, 2016. 3, 8

  37. [49]

    E. K. Tang, P. N. Suganthan, and X. Yao. An analysis of diversity measures. Machine Learning, 65(1):247–271, 2006. doi: 10.1007/s10994-006-9449 -2 4

  38. [50]

    Trabucco, K

    B. Trabucco, K. Doherty, M. Gurinas, and R. Salakhutdinov. Effective data augmentation with diffusion models. In Proceedings of the International Conference on Learning Representations, 2024. 3

  39. [51]

    Van der Maaten and G

    L. Van der Maaten and G. Hinton. Visualizing data using t-sne. Journal of Machine Learning Research, 9(11):2579–2605, 2008. 7

  40. [52]

    Y . Wang, S. Shen, and B. Y . Lim. RePrompt: Automatic prompt editing to refine ai-generative art towards precise expressions. In Proceedings of the CHI Conference on Human Factors in Computing Systems, pp. 23–32,

  41. [53]

    Z. Wang, Y . Huang, D. Song, L. Ma, and T. Zhang. PromptCharm: Text- to-image generation through multi-modal prompting and refinement. In Proceedings of the CHI Conference on Human Factors in Computing Systems, pp. 1–21, 2024. 2, 3

  42. [54]

    Z. Wen. Convergence of end-to-end training in deep unsupervised con- trastive learning. arXiv preprint arXiv:2002.06979, 2024. 5

  43. [55]

    doi: 10.1145/3544548.3581402 2

  44. [56]

    J. Xia, L. Huang, W. Lin, X. Zhao, J. Wu, Y . Chen, Y . Zhao, and W. Chen. Interactive visual cluster analysis by contrastive dimensionality reduction. IEEE Transactions on Visualization and Computer Graphics, 29(1):734– 744, 2022. doi: 10.1109/tvcg.2022.3209423 7

  45. [57]

    Xiang, X

    S. Xiang, X. Ye, J. Xia, J. Wu, Y . Chen, and S. Liu. Interactive correction of mislabeled training data. In 2019 IEEE Conference on Visual Analytics Science and Technology, pp. 57–68, 2019. doi: 10.1109/vast47406.2019. 8986943 5

  46. [58]

    S. Wold, K. Esbensen, and P. Geladi. Principal component analysis. Chemometrics and Intelligent Laboratory Systems, 2(1-3):37–52, 1987. doi: 10.1007/springerreference_84147 7

  47. [59]

    Y . Xu, L. Shang, J. Ye, Q. Qian, Y .-F. Li, B. Sun, H. Li, and R. Jin. Dash: Semi-supervised learning with dynamic thresholding. In International Conference on Machine Learning, pp. 11525–11536, 2021. 2

  48. [60]

    Yamaguchi and T

    S. Yamaguchi and T. Fukuda. On the limitation of diffusion models for synthesizing training datasets. In NeurIPS Workshop on Synthetic Data Generation with Generative AI, 2023. 1

  49. [61]

    X. Xie, X. Cai, J. Zhou, N. Cao, and Y . Wu. A semantic-based method for visualizing large image collections. IEEE Transactions on Visualization and Computer Graphics, 25(7):2362–2377, 2019. doi: 10.1109/TVCG. 2018.2835485 5

  50. [62]

    W. Yang, X. Ye, X. Zhang, L. Xiao, J. Xia, Z. Wang, J. Zhu, H. Pfister, and S. Liu. Diagnosing ensemble few-shot classifiers. IEEE Transactions on Visualization and Computer Graphics, 28(9):3292–3306, 2022. doi: 10 .1109/TVCG.2022.3182488 2

  51. [63]

    Y . Ye, S. Xiao, X. Zeng, and W. Zeng. ModalChorus: Visual probing and alignment of multi-modal embeddings via modal fusion map. arXiv preprint arXiv:2407.12315, 2024. doi: 10.1109/tvcg.2024.3456387 2, 5, 7

  52. [64]

    W. Yang, M. Liu, Z. Wang, and S. Liu. Foundation models meet visualiza- tions: Challenges and opportunities. Computational Visual Media, 2023. to be published. doi: 10.1007/s41095-023-0393-x 1

  53. [65]

    Zhang, D

    Y . Zhang, D. Zhou, B. Hooi, K. Wang, and J. Feng. Expanding small-scale datasets with guided imagination. In Proceedings of the Advances in Neural Information Processing Systems, pp. 76558–76618, 2023. 1, 3, 4, 9

  54. [66]

    Y . Zhou, W. Yang, J. Chen, C. Chen, Z. Shen, X. Luo, L. Yu, and S. Liu. Cluster-aware grid layout. IEEE Transactions on Visualization and Com- puter Graphics, 30(1):240–250, 2024. doi: 10.1109/TVCG.2023.3326934 5

  55. [67]

    Zhang, C

    Q. Zhang, C. Chen, Z. Liu, and Z. Tang. I-Adapt: Using iou adapter to improve pseudo labels in cross-domain object detection. In European Conference on Artificial Intelligence, pp. 57–64, 2024. 9

  56. [68]

    H. Zhu, L. Yang, J.-H. Yong, W. Zhang, and B. Wang. Distribution-aware data expansion with diffusion models. arXiv preprint arXiv:2403.06741,

  57. [70]

    Z.-H. Zhou, Y . Yu, and C. Qian. Evolutionary Learning: Advances in Theories and Algorithms. Springer Singapore, 2019. doi: 10.1007/978 -981-13-5956-9 6

  58. [2011]

    doi: 10.1109/HICSS.2011.339 8

  59. [2023]

    doi: 10.1145/3586183.3606777 2

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.