Pith. sign in

REVIEW 4 major objections 6 minor 54 references

Transmission Line Defect Detection Based on UAV Patrol Images and Vision-language Pretraining

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A three-stage pipeline that pretrains the detector backbone on transmission line image-text pairs, then progressively transfers it, lifts mAP50 on UAV patrol defect detection by 6.5 points over a ViTDet baseline.

desk verdict Competent domain-specific VLP application with internally consistent ablations, but the claimed gains rest on an unverified disjointness assumption between pretraining and test data. read the letter →

arxiv 2411.11370 v2 pith:5PLXCEI6 submitted 2024-11-18 cs.CV

Ke Zhang , Zhaoye Zheng , Yurong Guo , Jiacun Wang , Jiyuan Yang , Yangjie Xiao This is my paper · ORCID
classification cs.CV
keywords transmissionlinedefectdetectionUAVpatrolinspectionvision-languagepretrainingimage-textcontrastivelearningsamplerelationjudgmentdefect-normalcomparisonprogressivetransfersmall-object
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper claims that the accuracy of defect detection in UAV transmission line patrol images can be improved by pretraining the detector's image encoder with vision-language pretraining tailored to the transmission line domain, rather than relying only on the visual information in patrol images. It proposes VLP-TL, a pretraining method that combines image-text contrastive learning, sample relation judgment, and defect-normal comparison on a curated set of 23,391 transmission line image-text pairs, followed by a progressive transfer strategy that adds context-rich crops before fine-tuning the detector. On its private TLDD dataset, the resulting detector reaches 78.1% mAP50, 54.9% mAP75, and 47.2% mAP50:95, exceeding a ViTDet baseline by 6.5%, 2.6%, and 2.3% respectively. The intended significance is that linguistic information can compensate for defects that are small, occluded, or set in complex backgrounds, without changing the detector architecture or inference speed.

What carries the argument

The main machinery is the VLP-TL pretraining algorithm operating on a curated multimodal dataset of 23,391 transmission line image-text pairs. Three losses are combined: ITC retains the contrastive alignment between images and alt-texts; SRJ asks the model to classify the relation between two samples as same-type-same-status, same-type-different-status, or different-type; DNC compares features of defect and normal samples of the same component type with a binary similarity target. PTS then introduces instance-level images with context, generated by multiscale cropping around defects, and continues the same pretraining losses before the encoder is transferred to a ViTDet-style detection decoder.

What would settle it

Run the same pipeline with the TLDD split regrouped so that all images from a given patrol flight stay together in either training or testing; if the mAP50 improvement over ViTDet drops below the reported 6.5 percentage points, the original gain partly came from data overlap or label leakage. Alternatively, keep all settings fixed but replace the template-generated alt-texts with scrambled or category-free text; if mAP50 remains at 78.1%, the linguistic content is not the cause of the gain.

Watch

Extended reading notes

Core claim

The central discovery is that a domain-specific vision-language pretraining procedure, applied to an image encoder before transfer to a defect detector, improves detection accuracy on UAV transmission line patrol images. The paper reports that VLP-TL with SRJ, DNC, and ITC, followed by the PTS transition stage, yields mAP50 of 78.1%, mAP75 of 54.9%, and mAP50:95 of 47.2% on the TLDD dataset, compared with 71.6%, 52.3%, and 44.9% for ViTDet with an ImageNet-pretrained ViT-B backbone. The authors interpret this as evidence that multimodal knowledge helps overcome the insufficient defect-related visual information caused by imaging distance and shooting angle.

Load-bearing premise

The central claim rests on the assumption that the multimodal pretraining data are properly disjoint from the TLDD test set, and that the measured accuracy gains come from the image-text pretraining tasks rather than from the fact that the pretraining labels are the same categories the detector is trained to predict.

Editorial extensions

If this is right

  • If the reported gains hold, transmission line inspection systems can improve defect detection without changing detector parameters or inference speed, because VLP-TL and PTS only modify the training procedure.
  • The results suggest that domain-specific vision-language pretraining can work with a relatively small multimodal dataset (23,391 pairs), which would make the approach feasible for other power-system inspection scenarios.
  • The success of the transition stage implies that the gap between instance-level pretraining and full-scene detection can be partially closed with context-rich intermediate crops, a lesson that may transfer to other small-object detection tasks.
  • The ablation pattern indicates that contrastive learning must be retained as a base task; removing ITC while keeping SRJ and DNC drops mAP50 below the no-pretraining baseline.

Reading between the lines

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

  • The reported gains may partly reflect that pretraining and downstream detection share the same category annotation system; a leakage check that verifies the multimodal set is strictly disjoint from the TLDD test set would clarify how much of the improvement is real.
  • Because the alt-texts are created from predefined templates and category names, the linguistic signal is essentially a re-encoding of the labels; testing with human-written, richer descriptions would show whether additional text diversity helps.
  • The same three-stage pipeline could be tested on other aerial small-object domains, such as infrastructure or agricultural inspection, where the claim that text compensates for weak visual cues is directly checkable.
  • Repeated random splits and significance testing of the TLDD dataset would show whether the 2.3-6.5 percentage-point gaps are stable or within noise; the paper does not report such statistics.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes a three-stage training pipeline for transmission-line defect detection from UAV patrol images. Stage 1 (VLP-TL) pretrains a ViT-B image encoder and a text encoder on 23,391 transmission-line image-text pairs using image-text contrastive learning (ITC) plus two new tasks: sample relation judgment (SRJ) and defect-normal comparison (DNC). Stage 2 (PTS) adds an intermediate transition stage in which multiscale context crops from the TLDD dataset are mixed into the multimodal training data while the pretraining objectives are retained. Stage 3 fine-tunes a ViTDet-based detector on the TLDD dataset. The authors report that the resulting detector improves mAP50 by 6.5%, mAP75 by 2.6%, and mAP50:95 by 2.3% over a ViTDet baseline, and ablations attribute gains to both VLP-TL and PTS.

Significance. If the reported gains are attributable to the proposed method rather than to data overlap or test-set selection, this is a practically useful contribution for a domain where UAV patrol imagery is abundant but defect annotations are scarce. The paper has several strengths: the ablations in Tables III-V are internally consistent; the comparison against ViTDet is architecture-matched; the proposed PTS does not add parameters or inference cost; and the attention-map visualizations provide qualitative evidence of improved focus on defective regions. The main limitations are that all experiments use a private dataset, the test set is only about 183 images, no significance tests or multiple-seed results are reported, and the causal attribution to multimodal information depends on data-disjointness assumptions that are not verified.

major comments (4)
  1. [III-B / IV-A] The attribution of the reported gains to VLP-TL and PTS requires that no pretraining or transition data overlap with the TLDD test set. Section III-B states that PTS obtains instance-level context crops 'from the TLDD dataset' without restricting them to the training split, while Section IV-A only says TLDD is built from 'a part' of defect-containing images and the multimodal set from 'the remaining images.' If PTS crops or pretraining images come from the same flights or scenes as the test images, or are near-duplicates of them, the backbone is directly optimized on test-set content. Please state explicitly that all PTS crops come from the TLDD training split, describe how the TLDD/multimodal split was performed (e.g., flight-level or scene-level partition), and report a deduplication analysis between pretraining/PTS data and the TLDD test set.
  2. [III-A2 / Table IV / Table V] The two novel tasks SRJ and DNC construct supervision targets from the same category annotations (normal/defect component categories) that the downstream detector must predict, as shown in Eqs. (4), (9), and (10). The gain of VLP-TL over the ITC-only baseline in Table V (76.5 vs. 75.4 mAP50) could therefore reflect direct in-domain label supervision rather than multimodal alignment. Table IV also does not fully isolate the language contribution, because the image-classification baseline and the ITC baseline both derive from the same category annotations. Please add a control that uses the same images and the same category-derived targets without text, or that removes SRJ/DNC while keeping text, to demonstrate that the linguistic modality is the source of the improvement.
  3. [IV-C / Table II / Table V / Table VI] The headline improvements are computed on a private test set of about 183 images (10% of 1,830) with a single run and no error bars. Differences such as the 1.1 mAP50 gain from adding SRJ and DNC to ITC (Table V) and the 0.4 mAP50 difference between three and four PTS sizes (Table VI) are within plausible run-to-run variation for detection training. Please report results over multiple seeds with means and standard deviations or confidence intervals, and state the test-set size explicitly.
  4. [IV-D3 / Table VI] The number of PTS context sizes is selected by comparing downstream mAP on the TLDD test set, with Table VI peaking at three sizes. This makes the reported final numbers optimistic because the test set has already been used for hyperparameter selection. Please either use a held-out validation split for this choice or report both validation-based selection and final test performance.
minor comments (6)
  1. [Abstract / throughout] There are several typos and spacing errors: 'aimimg' should be 'aiming', 'pertaining' should be 'pretraining' in Sections I and IV, 'defection detection' should be 'defect detection', and 'UA V' should be 'UAV' throughout.
  2. [Fig. 6] The caption lists both '(e) A defect of foreign body' and '(e) A defect of bird nest and a defect of insulator bunch-drop'; the second should be labeled '(f)'.
  3. [III-B] The paper should clarify how category annotations and therefore SRJ/DNC targets are assigned to the instance-level context crops when these crops are mixed into the multimodal dataset during the transition stage.
  4. [Eq. (8)] Equation (8) uses the symbol ⊙ for what appears to be a matrix product; please clarify the notation and the dimensions of S.
  5. [IV-C / Table II] Please specify how the ViTDet baseline in Table II was initialized (ImageNet-pretrained or CN-CLIP-pretrained) and confirm that all compared detectors are trained under the same epochs, resolution, and data augmentation settings.
  6. [Fig. 8] The attention maps are qualitative; consider adding a quantitative localization or attention-focus metric to support the claim that the VLP-TL-pretrained encoder is more sensitive to components and defects.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the claimed gains are empirical and benchmarked against held-out TLDD test data with external baselines; residual concerns are data hygiene, not definitional circularity.

full rationale

The paper's causal chain is pretraining on a multimodal dataset, transferring the image encoder to a detector, and evaluating on a held-out TLDD test split. No equation or parameter is fitted to the reported test metric in a way that makes the result equivalent to the input by construction. SRJ and DNC use category annotations from the same annotation vocabulary as the detector, but the pretraining losses (Eqs. 4 and 9) are not the detection loss, and Table V shows non-forced outcomes (e.g., SRJ or DNC alone without ITC drops mAP50 to 41.8 while ITC+SRJ+DNC reaches 76.5), so the result is not a definitional reduction. The downstream comparison is against external, independently trained detectors (ViTDet, DETR-family, YOLO-family) and the TLDD evaluation uses a 9:1 train/test split, making the headline number a held-out measurement. The only substantial concern is that Section III-B obtains PTS context crops from the TLDD dataset without stating that they are restricted to the training split; if test images were used in the transition stage, the reported mAP would be inflated by leakage. That is a potential experimental-control flaw, not a derivational circularity, and cannot be scored as a definitional reduction without evidence. Self-citations (e.g., [8], [41], [54]) appear only in the related-work survey and are not load-bearing for the method's validity; the pretraining initialization and detection architecture come from independent sources (CN-CLIP [42], ViTDet [43]). Therefore the derivation chain is not circular.

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

The method introduces no new physical entities. It relies on a clean split between pretraining and detection images, on template-generated captions being informative, on CN-CLIP initialization, and on pretraining targets derived from the same category labels used downstream. The three free parameters (loss weights, DNC component weights, and the number of multiscale context sizes) are chosen by hand or by the observed peak in Table VI, and the loss weight values are not given.

free parameters (3)
  • Loss weights lambda_1, lambda_2, lambda_3 for ITC, SRJ, DNC (Eq. 12) = not specified
    Hand-chosen balancing of pretraining objectives; the central detection gains depend on this balance, and no sensitivity analysis is reported.
  • Number of multiscale context sizes in PTS = 3 (h1 x w1, h2 x w2, h3 x w3)
    Selected as the peak in Table VI, which reports mAP on the TLDD test set; this is post-hoc selection on the evaluation set.
  • Component-type weights alpha_c in DNC (Eq. 11) = data-dependent, not specified
    Set by sample count per component type; a design choice affecting the DNC loss balance.
assumptions (4)
  • domain assumption TLDD test images were not used in constructing the multimodal pretraining dataset (disjoint sets within the same patrol image pool).
    The paper states "select a part of defect-containing images" for TLDD and use "the remaining images" for the multimodal set (Section IV-A); the integrity of the central comparison depends on this split.
  • domain assumption The template-based alt-texts provide sufficient linguistic supervision for contrastive learning.
    Alt-texts are generated by filling predefined templates with category annotations (Section III-A1); the value of the language modality rests on these captions being informative and diverse.
  • domain assumption CN-CLIP pretrained weights are a valid initialization for both encoders.
    The method initializes from open-source CN-CLIP ViT-B-16 (Section IV-B); the reported gains are measured relative to this initialization.
  • ad hoc to paper SRJ and DNC targets derived from category annotations are valid proxies for structural defect knowledge.
    The pretraining tasks predict relations between samples using the same category labels that define the downstream detection classes (Eqs. 4-10), so the tasks are closely coupled to the evaluation target.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Transmission Line Defect Detection Based on UAV Patrol Images and Vision-language Pretraining." pith.science (2026). https://pith.science/paper/5PLXCEI6

@misc{pith2026241111370,
  author       = {Pith},
  title        = {Pith review of: Transmission Line Defect Detection Based on UAV Patrol Images and Vision-language Pretraining},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5PLXCEI6}},
  note         = {Machine review of arXiv:2411.11370}
}
read the original abstract

Unmanned aerial vehicle (UAV) patrol inspection has emerged as a predominant approach in transmission line monitoring owing to its cost-effectiveness. Detecting defects in transmission lines is a critical task during UAV patrol inspection. However, due to imaging distance and shooting angles, UAV patrol images often suffer from insufficient defect-related visual information, which has an adverse effect on detection accuracy. In this article, we propose a novel method for detecting defects in UAV patrol images, which is based on vision-language pretraining for transmission line (VLP-TL) and a progressive transfer strategy (PTS). Specifically, VLP-TL contains two novel pretraining tasks tailored for the transmission line scenario, aimimg at pretraining an image encoder with abundant knowledge acquired from both visual and linguistic information. Transferring the pretrained image encoder to the defect detector as its backbone can effectively alleviate the insufficient visual information problem. In addition, the PTS further improves transfer performance by progressively bridging the gap between pretraining and downstream defection detection. Experimental results demonstrate that the proposed method significantly improves defect detection accuracy by jointly utilizing multimodal information, overcoming the limitations of insufficient defect-related visual information provided by UAV patrol images.

Figures

Figures reproduced from arXiv: 2411.11370 by the authors.

Figure 1
Figure 1. An overview of the proposed method. a smoother transition between pretraining and defect detection. 3) After the pertaining and transition stage, we build a defect detector based on the image encoder and train it with the transmission line defect detection dataset (TLDD dataset). The experimental results show that our detector can effectively improve the defect detection accuracy in UAV patrol images. The rest of th… view at source ↗
Figure 3
Figure 3. The implementation of SRJ. i.e., same and different. In contrast, in the transmission line scenario, there are three probable relations between two sam￾ples, as shown in [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. The implementation of DNC, taking the case of grading rings as an [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figures from the paper (4 more)
Figure 5
Figure 5. Figure 5: The method to obtain instance-level images with context. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Data examples in the TLDD dataset. (a) A defect of grading ring [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Detection results of the proposed detector. (a) A small-sized insulator bunch-drop defect. (b) A damaged grading ring occluded by the tower. (c) A [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: Some images from the multimodal dataset and corresponding attention [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

54 extracted references · 35 canonical work pages

  1. [1]

    A review on state-of-the-art power line inspection techniques,

    L. Yang, J. Fan, Y . Liu, E. Li, J. Peng, and Z. Liang, “A review on state-of-the-art power line inspection techniques,” IEEE Transactions on Instrumentation and Measurement , vol. 69, no. 12, pp. 9350–9365, 2020

    work page 2020

  2. [2]

    Vision-based plane estimation and following for building inspection with autonomous uav,

    Y . Lyu, M. Cao, S. Yuan, and L. Xie, “Vision-based plane estimation and following for building inspection with autonomous uav,” IEEE Transactions on Systems, Man, and Cybernetics: Systems , 2023

    work page 2023

  3. [3]

    A bionic dynamic path planning algorithm of the micro uav based on the fusion of deep neural network optimization/filtering and hawk-eye vision,

    Z. Zhang, S. Wang, J. Chen, and Y . Han, “A bionic dynamic path planning algorithm of the micro uav based on the fusion of deep neural network optimization/filtering and hawk-eye vision,” IEEE Transactions on Systems, Man, and Cybernetics: Systems , vol. 53, no. 6, pp. 3728– 3740, 2023

    work page 2023

  4. [4]

    A survey of intelligent trans- mission line inspection based on unmanned aerial vehicle,

    Y . Luo, X. Yu, D. Yang, and B. Zhou, “A survey of intelligent trans- mission line inspection based on unmanned aerial vehicle,” Artificial Intelligence Review, vol. 56, no. 1, pp. 173–201, 2023

    work page 2023

  5. [5]

    In-yolo: Real- time detection of outdoor high voltage insulators using uav imaging,

    D. Sadykova, D. Pernebayeva, M. Bagheri, and A. James, “In-yolo: Real- time detection of outdoor high voltage insulators using uav imaging,” IEEE Transactions on Power Delivery , vol. 35, no. 3, pp. 1599–1601, 2019

    work page 2019

  6. [6]

    Hybrid knowledge r-cnn for transmission line multifitting detection,

    Y . Zhai, X. Yang, Q. Wang, Z. Zhao, and W. Zhao, “Hybrid knowledge r-cnn for transmission line multifitting detection,” IEEE Transactions on Instrumentation and Measurement , vol. 70, pp. 1–12, 2021

    work page 2021

  7. [7]

    Multi-fitting detec- tion on transmission line based on cascade reasoning graph network,

    Y . Zhai, Q. Wang, X. Yang, Z. Zhao, and W. Zhao, “Multi-fitting detec- tion on transmission line based on cascade reasoning graph network,” IEEE Transactions on Power Delivery , vol. 37, no. 6, pp. 4858–4868, 2022

    work page 2022

  8. [8]

    Transmission line key components and defects detection based on meta- learning,

    C. Dong, K. Zhang, Z. Xie, J. Wang, X. Guo, C. Shi, and Y . Xiao, “Transmission line key components and defects detection based on meta- learning,” IEEE Transactions on Instrumentation and Measurement , 2024

    work page 2024

Show all 54 references
  1. [9]

    Learning transferable visual models from natural language supervision,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark et al., “Learning transferable visual models from natural language supervision,” in International conference on machine learning . PMLR, 2021, pp. 8748–8763

    2021

  2. [10]

    Scaling up visual and vision-language representation learning with noisy text supervision,

    C. Jia, Y . Yang, Y . Xia, Y .-T. Chen, Z. Parekh, H. Pham, Q. Le, Y .-H. Sung, Z. Li, and T. Duerig, “Scaling up visual and vision-language representation learning with noisy text supervision,” in International conference on machine learning . PMLR, 2021, pp. 4904–4916

    2021

  3. [11]

    Winclip: Zero-/few-shot anomaly classification and segmentation,

    J. Jeong, Y . Zou, T. Kim, D. Zhang, A. Ravichandran, and O. Dabeer, “Winclip: Zero-/few-shot anomaly classification and segmentation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 19 606–19 616

    2023

  4. [12]

    Multimodality helps unimodality: Cross-modal few-shot learning with multimodal models,

    Z. Lin, S. Yu, Z. Kuang, D. Pathak, and D. Ramanan, “Multimodality helps unimodality: Cross-modal few-shot learning with multimodal models,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 19 325–19 337

    2023

  5. [13]

    Towards artificial general intelligence via a multimodal foundation model,

    N. Fei, Z. Lu, Y . Gao, G. Yang, Y . Huo, J. Wen, H. Lu, R. Song, X. Gao, T. Xiang et al., “Towards artificial general intelligence via a multimodal foundation model,” Nature Communications , vol. 13, no. 1, p. 3094, 2022

    2022

  6. [14]

    Learning to prompt for vision- language models,

    K. Zhou, J. Yang, C. C. Loy, and Z. Liu, “Learning to prompt for vision- language models,” International Journal of Computer Vision , vol. 130, no. 9, pp. 2337–2348, 2022

    2022

  7. [15]

    Cora: Adapting clip for open- vocabulary detection with region prompting and anchor pre-matching,

    X. Wu, F. Zhu, R. Zhao, and H. Li, “Cora: Adapting clip for open- vocabulary detection with region prompting and anchor pre-matching,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023, pp. 7031–7040

    2023

  8. [16]

    The unreasonable effectiveness of clip features for image captioning: an experimental analysis,

    M. Barraco, M. Cornia, S. Cascianelli, L. Baraldi, and R. Cucchiara, “The unreasonable effectiveness of clip features for image captioning: an experimental analysis,” in proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2022, pp. 4662–4670

    2022

  9. [17]

    Image as a foreign language: Beit pretraining for vision and vision-language tasks,

    W. Wang, H. Bao, L. Dong, J. Bjorck, Z. Peng, Q. Liu, K. Aggarwal, O. K. Mohammed, S. Singhal, S. Som et al. , “Image as a foreign language: Beit pretraining for vision and vision-language tasks,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni...

    2023

  10. [18]

    Improving clip training with language rewrites,

    L. Fan, D. Krishnan, P. Isola, D. Katabi, and Y . Tian, “Improving clip training with language rewrites,” Advances in Neural Information Processing Systems, vol. 36, pp. 35 544–35 575, 2023

    2023

  11. [19]

    Unified contrastive learning in image-text-label space,

    J. Yang, C. Li, P. Zhang, B. Xiao, C. Liu, L. Yuan, and J. Gao, “Unified contrastive learning in image-text-label space,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 19 163–19 173. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. ...

    2022

  12. [20]

    S-clip: Semi-supervised vision- language learning using few specialist captions,

    S. Mo, M. Kim, K. Lee, and J. Shin, “S-clip: Semi-supervised vision- language learning using few specialist captions,” Advances in Neural Information Processing Systems , vol. 36, pp. 61 187–61 212, 2023

    2023

  13. [21]

    Medclip: Contrastive learning from unpaired medical images and text,

    Z. Wang, Z. Wu, D. Agarwal, and J. Sun, “Medclip: Contrastive learning from unpaired medical images and text,” in Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing , 2022, pp. 3876–3887

    2022

  14. [22]

    Toward generalist anomaly detection via in-context residual learning with few-shot sample prompts,

    J. Zhu and G. Pang, “Toward generalist anomaly detection via in-context residual learning with few-shot sample prompts,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 17 826–17 836

    2024

  15. [23]

    Fash- ionvil: Fashion-focused vision-and-language representation learning,

    X. Han, L. Yu, X. Zhu, L. Zhang, Y .-Z. Song, and T. Xiang, “Fash- ionvil: Fashion-focused vision-and-language representation learning,” in European conference on computer vision. Springer, 2022, pp. 634–651

    2022

  16. [24]

    Imagenet: A large-scale hierarchical image database,

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, “Imagenet: A large-scale hierarchical image database,” in 2009 IEEE conference on computer vision and pattern recognition . Ieee, 2009, pp. 248–255

    2009

  17. [25]

    An accurate and real-time method of self-blast glass insulator location based on faster r-cnn and u-net with aerial images,

    Z. Ling, D. Zhang, R. C. Qiu, Z. Jin, Y . Zhang, X. He, and H. Liu, “An accurate and real-time method of self-blast glass insulator location based on faster r-cnn and u-net with aerial images,” CSEE Journal of Power and Energy Systems , vol. 5, no. 4, pp. 474–482, 2019

    2019

  18. [26]

    Detection and evaluation method of transmission line defects based on deep learning,

    H. Liang, C. Zuo, and W. Wei, “Detection and evaluation method of transmission line defects based on deep learning,” IEEE Access, vol. 8, pp. 38 448–38 458, 2020

    2020

  19. [27]

    An ultrasmall bolt defect detection method for transmission line inspection,

    P. Luo, B. Wang, H. Wang, F. Ma, H. Ma, and L. Wang, “An ultrasmall bolt defect detection method for transmission line inspection,” IEEE Transactions on Instrumentation and Measurement , vol. 72, pp. 1–12, 2023

    2023

  20. [28]

    Vision-language models for vision tasks: A survey,

    J. Zhang, J. Huang, S. Jin, and S. Lu, “Vision-language models for vision tasks: A survey,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024

    2024

  21. [29]

    Lasp: Text-to-text optimization for language-aware soft prompting of vision & language models,

    A. Bulat and G. Tzimiropoulos, “Lasp: Text-to-text optimization for language-aware soft prompting of vision & language models,” in Pro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023, pp. 23 232–23 241

    2023

  22. [30]

    Clip-adapter: Better vision-language models with feature adapters,

    P. Gao, S. Geng, R. Zhang, T. Ma, R. Fang, Y . Zhang, H. Li, and Y . Qiao, “Clip-adapter: Better vision-language models with feature adapters,” International Journal of Computer Vision , vol. 132, no. 2, pp. 581–595, 2024

    2024

  23. [31]

    Graphadapter: Tuning vision-language models with dual knowledge graph,

    X. Li, D. Lian, Z. Lu, J. Bai, Z. Chen, and X. Wang, “Graphadapter: Tuning vision-language models with dual knowledge graph,” Advances in Neural Information Processing Systems , vol. 36, pp. 13 448–13 466, 2023

    2023

  24. [32]

    Finetune like you pretrain: Improved finetuning of zero-shot vision models,

    S. Goyal, A. Kumar, S. Garg, Z. Kolter, and A. Raghunathan, “Finetune like you pretrain: Improved finetuning of zero-shot vision models,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 19 338–19 347

    2023

  25. [33]

    Few-shot electrical equipment image recognition method based on an improved two-stage fine-tuning approach,

    J. Wu, J. Zeng, Y . Zhou, Y . Zhang, and Y . Zhang, “Few-shot electrical equipment image recognition method based on an improved two-stage fine-tuning approach,” The Journal of Engineering , vol. 2023, no. 9, p. e12313, 2023

    2023

  26. [34]

    Defect detection method for key area guided transmission line components based on knowledge distillation,

    Z. Zhao, X. Lv, Y . Xi, and S. Miao, “Defect detection method for key area guided transmission line components based on knowledge distillation,” Frontiers in Energy Research, vol. 11, p. 1287024, 2023

    2023

  27. [35]

    A defect detection method based on bc-yolo for transmission line components in uav remote sensing images,

    W. Bao, X. Du, N. Wang, M. Yuan, and X. Yang, “A defect detection method based on bc-yolo for transmission line components in uav remote sensing images,” Remote Sensing, vol. 14, no. 20, p. 5176, 2022

    2022

  28. [36]

    A robust faster r-cnn model with feature enhancement for rust detection of transmission line fitting,

    Z. Guo, Y . Tian, and W. Mao, “A robust faster r-cnn model with feature enhancement for rust detection of transmission line fitting,” Sensors, vol. 22, no. 20, p. 7961, 2022

    2022

  29. [37]

    Ecc-rcnn: An efficient and high-accuracy object detection framework for transmission line defect identification,

    Y . Li, Y . Xu, W. Sun, Q. Qian, Z. Li, and X. Jiang, “Ecc-rcnn: An efficient and high-accuracy object detection framework for transmission line defect identification,” IET Smart Grid , vol. 7, no. 1, pp. 28–37, 2024

    2024

  30. [38]

    Pkamnet: a transmission line insulator parallel-gap fault detection network based on prior knowledge transfer and attention mechanism,

    S. Hao, B. An, X. Ma, X. Sun, T. He, and S. Sun, “Pkamnet: a transmission line insulator parallel-gap fault detection network based on prior knowledge transfer and attention mechanism,” IEEE Transactions on Power Delivery, vol. 38, no. 5, pp. 3387–3397, 2023

    2023

  31. [39]

    Attention-based multimodal image feature fusion module for transmission line detection,

    H. Choi, J. P. Yun, B. J. Kim, H. Jang, and S. W. Kim, “Attention-based multimodal image feature fusion module for transmission line detection,” IEEE Transactions on Industrial Informatics , vol. 18, no. 11, pp. 7686– 7695, 2022

    2022

  32. [40]

    Ptsrdet: End- to-end super-resolution and object-detection approach for small defect detection of power transmission lines,

    S. Falahatnejad, A. Karami, and H. Nezamabadi-pour, “Ptsrdet: End- to-end super-resolution and object-detection approach for small defect detection of power transmission lines,” Computers and Electrical Engi- neering, vol. 124, p. 110374, 2025

    2025

  33. [41]

    Transmission line component defect detection based on uav patrol images: A self- supervised hc-vit method,

    K. Zhang, R. Zhou, J. Wang, Y . Xiao, X. Guo, and C. Shi, “Transmission line component defect detection based on uav patrol images: A self- supervised hc-vit method,” IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2024

    2024

  34. [42]

    Chinese clip: Contrastive vision-language pretraining in chinese,

    A. Yang, J. Pan, J. Lin, R. Men, Y . Zhang, J. Zhou, and C. Zhou, “Chinese clip: Contrastive vision-language pretraining in chinese,” arXiv preprint arXiv:2211.01335, 2022

    2022 arXiv

  35. [43]

    Exploring plain vision transformer backbones for object detection,

    Y . Li, H. Mao, R. Girshick, and K. He, “Exploring plain vision transformer backbones for object detection,” in European conference on computer vision . Springer, 2022, pp. 280–296

    2022

  36. [44]

    Faster r-cnn: Towards real-time object detection with region proposal networks,

    S. Ren, K. He, R. Girshick, and J. Sun, “Faster r-cnn: Towards real-time object detection with region proposal networks,” IEEE transactions on pattern analysis and machine intelligence, vol. 39, no. 6, pp. 1137–1149, 2016

    2016

  37. [45]

    Cascade r-cnn: High quality object detection and instance segmentation,

    Z. Cai and N. Vasconcelos, “Cascade r-cnn: High quality object detection and instance segmentation,” IEEE transactions on pattern analysis and machine intelligence, vol. 43, no. 5, pp. 1483–1498, 2019

    2019

  38. [46]

    Sparse r-cnn: An end-to-end framework for object detection,

    P. Sun, R. Zhang, Y . Jiang, T. Kong, C. Xu, W. Zhan, M. Tomizuka, Z. Yuan, and P. Luo, “Sparse r-cnn: An end-to-end framework for object detection,” IEEE transactions on pattern analysis and machine intelligence, vol. 45, no. 12, pp. 15 650–15 664, 2023

    2023

  39. [47]

    Focal loss for dense object detection,

    T.-Y . Lin, P. Goyal, R. Girshick, K. He, and P. Doll ´ar, “Focal loss for dense object detection,” in Proceedings of the IEEE international conference on computer vision , 2017, pp. 2980–2988

    2017

  40. [48]

    Rtmdet: An empirical study of designing real-time object detectors,

    C. Lyu, W. Zhang, H. Huang, Y . Zhou, Y . Wang, Y . Liu, S. Zhang, and K. Chen, “Rtmdet: An empirical study of designing real-time object detectors,” arXiv preprint arXiv:2212.07784 , 2022

    2022 arXiv

  41. [49]

    Ultralytics YOLO,

    G. Jocher, J. Qiu, and A. Chaurasia, “Ultralytics YOLO,” Jan. 2023. [Online]. Available: https://github.com/ultralytics/ultralytics

    2023

  42. [50]

    End-to-end object detection with transformers,

    N. Carion, F. Massa, G. Synnaeve, N. Usunier, A. Kirillov, and S. Zagoruyko, “End-to-end object detection with transformers,” in European conference on computer vision . Springer, 2020, pp. 213– 229

    2020

  43. [51]

    Deformable detr: De- formable transformers for end-to-end object detection,

    X. Zhu, W. Su, L. Lu, B. Li, X. Wang, and J. Dai, “Deformable detr: De- formable transformers for end-to-end object detection,” in International Conference on Learning Representations , 2021

    2021

  44. [52]

    Conditional detr for fast training convergence,

    D. Meng, X. Chen, Z. Fan, G. Zeng, H. Li, Y . Yuan, L. Sun, and J. Wang, “Conditional detr for fast training convergence,” in Proceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 3651– 3660

    2021

  45. [53]

    Insulator defect detection under extreme weather based on synthetic weather algorithm and improved yolov7,

    Y . Yang, S. Yang, C. Li, Y . Wang, X. Pi, Y . Lu, and R. Wu, “Insulator defect detection under extreme weather based on synthetic weather algorithm and improved yolov7,”High Voltage, vol. 10, no. 1, pp. 69–77, 2025

    2025

  46. [54]

    Pa-detr: End-to-end visually indistinguishable bolt defects detection method based on transmission line knowledge reasoning,

    K. Zhang, W. Lou, J. Wang, R. Zhou, X. Guo, Y . Xiao, C. Shi, and Z. Zhao, “Pa-detr: End-to-end visually indistinguishable bolt defects detection method based on transmission line knowledge reasoning,” IEEE Transactions on Instrumentation and Measurement , vol. 72, pp. 1–14, 2023

    2023

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.