Pith. sign in

REVIEW 5 major objections 5 minor 26 references

SAViL-Det: Semantic-Aware Vision-Language Model for Multi-Script Text Detection

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

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

Pith's one-line read A text detector that feeds CLIP language prompts into pixel-level features reports top F-scores of 84.8% on MLT-2019 and 90.2% on CTW1500.

desk verdict The architecture idea is plausible, but the 'no text' ablation on CTW1500 exactly duplicates TextSnake's published numbers, so the paper's central claim lacks credible support as it stands. read the letter →

arxiv 2507.20188 v1 pith:NKQROKZ4 submitted 2025-07-27 cs.CV

Mohammed-En-Nadhir Zighem , Abdenour Hadid This is my paper
classification cs.CV
keywords scenetextdetectionvision-languagemodelCLIPmulti-lingualcurvedcross-modalattentioncontrastivelearning
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

SAViL-Det is a proposal for detecting text in natural images by treating the text prompt as a first-class input rather than an optional helper. The paper claims that a decoder which first applies self-attention to visual features and then cross-attention with features from a frozen CLIP text encoder lets semantic information from the prompt propagate into pixel-level features, and that a text-to-pixel contrastive loss makes those features align with the prompt. On two benchmark settings the paper reports F-scores of 84.8% on the multi-lingual MLT-2019 set and 90.2% on the curved-text CTW1500 set, ahead of the comparison methods in its tables. This matters because multi-script and curved text are exactly the cases where visual shape alone is ambiguous, so a detector that can use language context promises to be more reliable in the wild.

What carries the argument

The load-bearing component is the language-vision decoder: a transformer layer stack in which multi-head self-attention runs over visual features, multi-head cross-attention takes keys and values from the frozen CLIP text tokens, and a feed-forward MLP with residual connections outputs multi-modal features. Around it sit the AFPN multi-scale fusion, which combines non-adjacent feature levels with learnable spatial weights, and a text/image projector that upsamples the decoder output to image resolution and applies a sigmoid contrastive loss between the global text vector and each pixel vector. Together these make the prompt's semantic content able to reshape pixel-level features before the detection head.

What would settle it

Re-run the 'No Textual Features' configuration on CTW1500 with the same training data and schedule; if the measured F-score is not close to the reported 75.6, the 14.6-point gap attributed to language input is not supported.

Watch

Extended reading notes

Core claim

The paper's central claim is that textual semantics are not just side information but the active ingredient that improves detection. Concretely, SAViL-Det encodes three script-agnostic prompts with the frozen CLIP text encoder, fuses multi-scale visual features from a CLIP ResNet-50 backbone with an Asymptotic Feature Pyramid Network, and then runs a decoder whose cross-modal attention lets each visual token attend to text-token features. A contrastive loss over the global text embedding and projected pixel embeddings pushes text pixels close to the prompt and background far from it. The paper reports that this pipeline reaches 84.8% F on MLT-2019 and 90.2% F on CTW1500, and that removing the textual features lowers the F-score by 3.4 points on MLT-2019 and from 90.2 to 75.6 on CTW1500.

Load-bearing premise

The load-bearing premise is that the 'No Textual Features' row was actually measured rather than carried over from a published baseline, since on CTW1500 its recall, precision, and F-score are exactly TextSnake's published numbers.

Editorial extensions

If this is right

  • If the reported scores hold, a single prompt-conditioned model can outperform dedicated multi-lingual and curved-text detectors without per-script specialisation.
  • The decoder-depth ablation indicates the benefit comes from exactly three transformer layers; going to four or five layers degrades F-scores on both datasets.
  • The prompt-ablation numbers imply that language input is worth 3.4 F-points on MLT-2019 and about 14.6 points on CTW1500, meaning curved text is where semantic guidance pays off most.
  • Because the prompt list names the scripts to detect, the same model outputs script labels as well as boxes on MLT-2019.

Reading between the lines

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

  • A natural next test, not run in the paper, is to change the prompt's script names or replace them with nonsense words; if the F-score moves little, the gain might come from conditioning in general rather than from semantic script knowledge.
  • The contrastive loss treats text-to-pixel similarity with a single sigmoid; the paper does not specify how negative pairs are sampled, so varying the negative sampling rule (other text instances vs background) is a testable knob that could explain the CTW1500 precision jump.
  • If the decoder truly propagates token-level semantics, the architecture should extend to text spotting (reading the detected words) by conditioning on the actual transcription instead of generic prompts, a possibility the paper lists as future work.
  • The surprising CTW1500 no-text result, which matches TextSnake's published row digit for digit, should make readers treat the ablation gap as provisional rather than final; reproducing that row is the cheapest way to stress-test the main claim.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper proposes SAViL-Det, a scene text detection method that combines a frozen CLIP text encoder, a trainable ResNet-50 image encoder, an Asymptotic Feature Pyramid Network (AFPN), a transformer-based cross-modal decoder, and a text-to-pixel contrastive loss. The authors report state-of-the-art F-scores of 84.8% on MLT2019 and 90.2% on CTW1500, and they attribute these gains to the language-vision decoder and the text-to-pixel contrastive learning mechanism. The paper includes comparisons with prior methods, two ablation studies (decoder depth and removal of textual features), and qualitative examples.

Significance. The idea of injecting semantic text prompts into a multi-script text detector is timely and potentially useful, especially for handling curved and multilingual text where visual cues alone are insufficient. The paper is clearly organized and the experimental comparison covers a reasonable set of recent methods. However, as written, the method is not fully specified: the detection head and its training loss are never described, and the contrastive loss in Eq. (7) is not precisely defined. More seriously, the key ablation that is supposed to justify the textual component, Table 5, contains a row on CTW1500 that exactly duplicates the published TextSnake numbers from Table 3, and the accompanying text cites a full-model F-score of 89.6% that contradicts the 90.2% reported elsewhere. These issues mean that the central causal claim, that the language component is essential, is not currently supported by the evidence presented.

major comments (5)
  1. [Sections 3.3 and 3.4] The detection head and the actual training objective for text detection are never described. The paper reports detection F-scores in Section 4.4, but no equation or text explains how the aligned feature map produced in Section 3.4 is converted into text instance polygons, nor what loss (segmentation, regression, or otherwise) supervises that prediction. Without this information, the proposed architecture is incomplete and the reported performance cannot be reproduced or attributed to the proposed components.
  2. [Section 3.4, Eq. (7)] The sets P and N in the contrastive loss are not defined. The text says these are 'positive and negative pairs in the ground truth', but it never specifies which pixels are positive (all text pixels, per-instance pixels, or something else), how the single global text feature z_t is matched to multiple positive pixel features, or how negative pixels are sampled. Since the text-to-pixel contrastive learning is presented as a core contribution, this omission is load-bearing and prevents the reader from understanding or reimplementing the objective.
  3. [Section 4.5, Table 5 vs. Table 3] In the CTW1500 ablation, the 'No textual features' row reports R=85.3, P=67.9, F=75.6, which is exactly the same R/P/F reported for TextSnake in Table 3, including the unusual precision value of 67.9. Moreover, the paragraph interpreting this ablation states the drop is from 89.6% to 75.6%, although the full model in Table 3 has F=90.2%. The exact match to a different method's published numbers strongly suggests that this row was carried over from the baseline table rather than being an independently measured variant of the proposed model. Since this ablation is the only direct evidence presented that textual features are essential, the central causal claim is unsupported.
  4. [Section 3.2, Eq. (4)] The adaptive spatial fusion operation is written as F_v = α·F_v0 + β·F_v1 + γ·F_v3, with learnable weights α, β, γ, δ that sum to 1, but only three terms appear in the equation and δ is never defined, while F_v2 is omitted. This is inconsistent with Eqs. (1)–(3), where F_v1, F_v2, and F_v3 are computed sequentially. As written, the fusion mechanism is not well-defined and the reader cannot tell which features are actually combined.
  5. [Sections 4.3 and 4.5] Model selection appears to be performed directly on the test sets. The decoder-depth ablation in Table 4 reports F-scores on the MLT2019 test set and the CTW1500 test set for depths 2 through 5, and the authors conclude that 3 layers is optimal based on these test-set numbers; no validation split or cross-validation is mentioned. Similarly, the three prompt templates in Table 1 are selected without any prompt-level ablation. Selecting hyperparameters on the test set inflates the reported results and weakens the validity of the state-of-the-art comparison, even though the final metrics themselves are computed against external benchmarks.
minor comments (5)
  1. [Section 3.2, Eq. (1)] The text says '1× upsampling' but F1 is at a lower resolution than F0, so the upsampling factor should be 2×; please correct this typo.
  2. [Section 3.3] The notation is inconsistent: the paper writes 'Let Ft ∈ R^(h×w)×d and Fv ∈ R^(N×C)', but earlier Ft is defined as R^(L×C) for text tokens and Fv as a visual feature map. Assigning spatial dimensions (h×w) to textual features is confusing; please unify the notation.
  3. [Figure 3 caption] The caption says 'the decoder takes visual features Ft' but Ft denotes textual features in the main text; the variable should be Fv or the caption should be corrected.
  4. [Section 4.5] There is a typo: 'descrease' should be 'decrease'.
  5. [General] No code, checkpoints, or reproducibility statement is provided. Given the missing implementation details and the numerical inconsistency in Table 5, a reproducibility statement with links or logs is essential for any future revision.

Circularity Check

2 steps flagged · score 4.0 of 10

The CTW1500 'No textual features' ablation exactly duplicates TextSnake's published R/P/F, so the claim that language input is essential lacks independent measurement; decoder depth is also tuned on the target test sets.

  1. renaming known result [Section 4.5, Table 5 (Impact of text prompt features), compared with Section 4.4, Table 3 (TextSnake row)]
    "Table 5: Impact of language input ... 'No' row: 'No 80.2 82.7 81.4 85.3 67.9 75.6'; Table 3: 'TextSnake[9] 85.3 67.9 75.6'; text: 'On CTW1500, the decrease was even more significant —14% (89.6% to 75.6%).'"

    The CTW1500 'No' row is numerically identical to TextSnake's published result (R=85.3, P=67.9, F=75.6), down to the unusual precision value. TextSnake is a different, purely visual architecture, yet the paper presents this row as its own model with text input removed from the decoder. Presenting a published baseline's numbers as the proposed model's own no-text ablation is renaming a known result rather than measuring it. The paper then uses this row as its only direct evidence that 'the language component in our cross-modal decoder is essential'. The paragraph even computes the drop as '14% (89.6% to 75.6%)' while Table 3 reports the full model at 90.2, showing internal inconsistency. Without code, checkpoints, or logs, the claimed ablation is unverifiable.

  2. other [Section 4.5, 'Vision-language decoder's depth' and Table 4]
    "We evaluated the recall at different Intersection-over-Union thresholds (0.50, 0.60, 0.70, 0.80, and 0.90) on the MLT2019 and CTW1500 datasets. ... The findings indicate that 3 transformer layers offer an optimal balance for the vision-language decoder in our SAViL-Det method."

    The decoder depth is selected by comparing F-scores on the same MLT2019 and CTW1500 test sets that later serve as the headline SOTA comparison (84.8% and 90.2%). Choosing the architecture that maximizes the target metric on the test set, then reporting that test result as SOTA, is a mild selection circularity: the reported configuration is not an independent prediction but the best of several configurations scored directly on the target test data. This is separate from the external evaluative basis of the headline numbers but weakens the causal interpretation of the final comparison.

full rationale

The headline F-scores (84.8% on MLT2019, 90.2% on CTW1500) are measured against external benchmarks and are therefore not definitionally circular; there is no load-bearing self-citation chain, since the paper does not cite prior work by its own authors. The dominant problem is the 'No textual features' ablation: the CTW1500 'No' row in Table 5 reproduces TextSnake's exact published R/P/F (85.3/67.9/75.6) from Table 3, and the interpreting paragraph cites a full-model F-score of 89.6 while Table 3 reports 90.2. If that row is not an independently measured run, then the only direct evidence that textual features are essential is an imported baseline relabeled as the proposed model's own configuration. Additionally, decoder depth (3 layers) is selected by maximizing F-scores on the same MLT2019 and CTW1500 test sets that later serve as the SOTA comparison, a mild selection circularity. These issues undermine the causal story of the paper but do not make the externally reported F-scores circular by construction, so the overall score is 4 rather than 6 or higher.

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

The model relies on pretrained CLIP, AFPN, and standard transformer components, all taken from prior literature. The central claim does not introduce new physical entities; it depends on the correctness and transferability of these pretrained models, the completeness of the method specification (which is lacking), and the integrity of the experimental measurements, one of which appears duplicated from a baseline paper.

free parameters (2)
  • Text prompt templates = P1: 'Detect Any text in the image.'; P2: 'Where is text located in the scene?'; P3: 'Detect Any text in the scene.'
    Hand-crafted prompts chosen by the authors; no ablation across prompt choices is reported, so their contribution to the F-scores is not independently quantified.
  • Vision-language decoder depth = 3 layers
    Selected via ablation on MLT2019 and CTW1500 (Table 4) using F-scores that appear to come from the same test sets used for the final SOTA claim; this is post-hoc model selection on the benchmarks.
assumptions (4)
  • domain assumption CLIP provides transferable semantic alignment between text prompts and visual features.
    Invoked in Sec 3.1: the image and text encoders are taken from a pretrained CLIP model; the method's success depends on this alignment.
  • domain assumption AFPN fusion of non-adjacent feature levels preserves fine details and semantic information better than standard FPN.
    Adopted in Sec 3.2 based on [18]; the paper provides no independent validation of this claim for text detection.
  • domain assumption MLT-2019 and CTW1500 test annotations are correct and the standard evaluation protocol (with '###' ignored) is followed.
    Sec 4.2 describes the protocol; the reported F-scores rest on this assumption, and no validation split is mentioned for model selection.
  • ad hoc to paper The contrastive loss with sigmoid over pixel similarities is a valid objective for text-to-pixel alignment.
    Introduced in Sec 3.4 without a derivation or comparison to alternative objectives; positive and negative sets are undefined.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SAViL-Det: Semantic-Aware Vision-Language Model for Multi-Script Text Detection." pith.science (2026). https://pith.science/paper/NKQROKZ4

@misc{pith2026250720188,
  author       = {Pith},
  title        = {Pith review of: SAViL-Det: Semantic-Aware Vision-Language Model for Multi-Script Text Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NKQROKZ4}},
  note         = {Machine review of arXiv:2507.20188}
}
read the original abstract

Detecting text in natural scenes remains challenging, particularly for diverse scripts and arbitrarily shaped instances where visual cues alone are often insufficient. Existing methods do not fully leverage semantic context. This paper introduces SAViL-Det, a novel semantic-aware vision-language model that enhances multi-script text detection by effectively integrating textual prompts with visual features. SAViL-Det utilizes a pre-trained CLIP model combined with an Asymptotic Feature Pyramid Network (AFPN) for multi-scale visual feature fusion. The core of the proposed framework is a novel language-vision decoder that adaptively propagates fine-grained semantic information from text prompts to visual features via cross-modal attention. Furthermore, a text-to-pixel contrastive learning mechanism explicitly aligns textual and corresponding visual pixel features. Extensive experiments on challenging benchmarks demonstrate the effectiveness of the proposed approach, achieving state-of-the-art performance with F-scores of 84.8% on the benchmark multi-lingual MLT-2019 dataset and 90.2% on the curved-text CTW1500 dataset.

Figures

Figures reproduced from arXiv: 2507.20188 by the authors.

Figure 1
Figure 1. Our method versus text detection identification systems. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. SAViL-Det architecture. Given an input image and a text prompt, our [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The decoder takes visual features Ft ∈ R N×C and textual features Ft, applying multi-head self-attention (MHSA) to update the visual features, then multi-head cross-attention (MHCA) to integrate semantic information. The final multi-modal features Fc are produced via a multi-layer perceptron (MLP) with residual connections. where zt ∈ R D and zv ∈ R N×D, with N = H 4 × W 4 . Here, Upsample refers to a 4× upsampling … view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Visualization of text detection results on MLT2019 dataset (Top) and [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

26 extracted references · 25 canonical work pages

  1. [21]

    In: CVPR

    Yu, W., Liu, Y., Hua, W., Jiang, D., Ren, B., Bai, X.: Turning a clip model into a scene text detector. In: CVPR. pp. 6978–6988 (2023)

    work page 2023

  2. [17]

    In: European Conference on Computer Vision

    Xue, C., Zhang, W., Hao, Y., Lu, S., Torr, P.H., Bai, S.: Language matters: A weakly supervised vision-language pre-training approach for scene text detection and spotting. In: European Conference on Computer Vision. Springer (2022)

    work page 2022

  3. [1]

    In: Proc

    Baek, Y., Lee, B., Han, D., Yun, S., Lee, H.: Character region awareness for text detection. In: Proc. CVPR. pp. 9365–9374 (2019)

    work page 2019

  4. [2]

    IET Image Processing16(10), 1234–1245 (2022)

    Blanco-Medina, R., et al.: A survey on methods, datasets and implementations for scene text spotting. IET Image Processing16(10), 1234–1245 (2022)

    work page 2022

  5. [3]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Bu, Q., Park, S., Khang, M., Cheng, Y.: Srformer: Text detection transformer with incorporated segmentation and regression. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 38, pp. 855–863 (2024)

    work page 2024

  6. [4]

    Deng, D., Liu, H., Li, X., Cai, D.: Pixellink: Detecting scene text via instance segmentation (2018), https://arxiv.org/abs/1801.01315

    work page Pith review arXiv 2018

  7. [5]

    In: 2013 12th ICDAR

    Karatzas, D., Shafait, F., Uchida, S., Iwamura, M., i Bigorda, L.G., Mestre, S.R., Mas, J., Mota, D.F., Almazan, J.A., De Las Heras, L.P.: Icdar 2013 robust reading competition. In: 2013 12th ICDAR. pp. 1484–1493. IEEE (2013) 12 Zighem, Hadid

    work page 2013

  8. [6]

    https://github.com/facebookresearch/xformers (2022)

    Lefaudeux, B., Massa, F., Liskovich, D., Xiong, W., Caggiano, V., Naren, S., Xu, M., Hu, J., Tintore, M., Zhang, S., Labatut, P., Haziza, D., Wehrstedt, L., Reizen- stein, J., Sizov, G.: xformers: A modular and hackable transformer modelling li- brary. https://github.com/facebookresearch/xformers (2022)

    work page 2022

Show all 26 references
  1. [7]

    TPAMI45(1), 919–931 (2022)

    Liao, M., Zou, Z., Wan, Z., Yao, C., Bai, X.: Real-time scene text detection with differentiable binarization and adaptive scale fusion. TPAMI45(1), 919–931 (2022)

    2022

  2. [8]

    Pattern Recognition90 (2019)

    Liu, Y., Jin, L., Zhang, S., Luo, C., Zhang, S.: Curved scene text detection via transverse and longitudinal sequence connection. Pattern Recognition90 (2019)

    2019

  3. [9]

    In: ECCV

    Long, S., Ruan, J., Zhang, W., He, X., Wu, W., Yao, C.: Textsnake: A flexible representation for detecting text of arbitrary shapes. In: ECCV. pp. 20–36 (2018)

    2018

  4. [10]

    In: ICDAR

    Nayef, N., Patel, Y., Busta, M., Chowdhury, P.N., Karatzas, D., Khlif, W., Matas, J.,Pal,U.,Burie,J.C.,Liu,C.l.,etal.:Icdar2019robustreadingchallengeonmulti- lingual scene text detection and recognition. In: ICDAR. pp. 1582–1587 (2019)

    2019

  5. [11]

    In: ICCV (2021)

    Pan, Y., Huang, J., Chen, K., Gong, T., Liu, C., Bai, X., Liu, W.: Fast: Faster arbitrarily-shaped text detector with minimalist kernel representation. In: ICCV (2021)

    2021

  6. [12]

    In: ICML 2021

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

    2021

  7. [13]

    arXiv preprint arXiv:2006.04305 (2020)

    Raisi, Z., Naiel, M.A., Fieguth, P., Wardell, S., Zelek, J.: Text detection and recog- nition in the wild: A review. arXiv preprint arXiv:2006.04305 (2020)

    2020 arXiv

  8. [14]

    In: CVPR

    Song, S., Wan, J., Yang, Z., Tang, J., Cheng, W., Bai, X., Yao, C.: Vision-language pre-training for boosting scene text detectors. In: CVPR. pp. 15681–15691 (2022)

    2022

  9. [15]

    In: CVPR

    Wang, W., Xie, E., Li, X., Hou, W., Lu, T., Yu, G., Shao, S.: Shape robust text de- tection with progressive scale expansion network. In: CVPR. pp. 9336–9345 (2019)

    2019

  10. [16]

    In: CVPR

    Wang,W.,Xie,E.,Song,X.,Zang,Y.,Wang,W.,Lu,T.,Yu,G.,Shen,C.:Efficient and accurate arbitrary-shaped text detection with pixel aggregation network. In: CVPR. pp. 8440–8449 (2019)

    2019

  11. [18]

    In: 2023 IEEE International Conference on Systems, Man, and Cybernetics (SMC)

    Yang, G., Lei, J., Zhu, Z., Cheng, S., Feng, Z., Liang, R.: Afpn: Asymptotic feature pyramid network for object detection. In: 2023 IEEE International Conference on Systems, Man, and Cybernetics (SMC). pp. 2184–2189. IEEE (2023)

    2023

  12. [19]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Ye, M., Zhang, J., Zhao, S., Liu, J., Du, B., Tao, D.: Dptext-detr: Towards better scene text detection with dynamic points in transformer. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 37, pp. 3241–3249 (2023)

    2023

  13. [20]

    Ye, M., Zhang, J., Zhao, S., Liu, J., Liu, T., Du, B., Tao, D.: Deepsolo++: Let transformer decoder with explicit points solo for multilingual text spotting (2023)

    2023

  14. [22]

    arXiv preprint arXiv:2308.12817 (2023)

    Zeng, Y.X., Hsieh, J.W., Li, X., Chang, M.C.: Mixnet: toward accurate detection of challenging scene text in the wild. arXiv preprint arXiv:2308.12817 (2023)

    2023 arXiv

  15. [23]

    In: CVPR (2019)

    Zhang, C., Liang, B., Huang, Z., En, M., Han, J., Ding, E., Ding, X.: Look more than once: An accurate detector for text of arbitrary shapes. In: CVPR (2019)

    2019

  16. [24]

    In: Proc

    Zhang, X., Su, Y., Tripathi, S., Tu, Z.: Text spotting transformers. In: Proc. IEEE conference on computer vision and pattern recognition. pp. 9519–9528 (2022)

    2022

  17. [25]

    In: CVPR

    Zhou, X., Yao, C., Wen, H., Wang, Y., Zhou, S., He, W., Liang, J.: East: an efficient and accurate scene text detector. In: CVPR. pp. 5551–5560 (2017)

    2017

  18. [26]

    In: CVPR

    Zhu, Y., Chen, J., Liang, L., Kuang, Z., Jin, L., Zhang, W.: Fourier contour em- bedding for arbitrary-shaped text detection. In: CVPR. pp. 3123–3131 (2021)

    2021

Pith tools

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