Pith. sign in

REVIEW 3 major objections 5 minor 60 references

JVLGS: Joint Vision-Language Gas Leak Segmentation

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read JVLGS claims a joint vision-language video model with false-positive cleanup outperforms prior gas-leak segmentation methods on two benchmarks.

desk verdict The SimGas win is mostly a post-processing artifact tuned on the test set, so the central 'outperforms SOTA' claim is not yet established. read the letter →

arxiv 2508.19485 v1 pith:YTMASGFN submitted 2025-08-27 cs.CV

Xinlong Zhao , Qixiang Pang , Shan Du This is my paper · ORCID
classification cs.CV
keywords gasleaksegmentationvision-languagemodelinfraredvideofew-shotobjecttemporal-spatialmodulefalsepositivesuppressiontext-guided
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

JVLGS is a video segmentation model that takes short text prompts (such as "White Steam" and "Billowing Smoke") together with infrared video clips and returns a per-frame mask of the gas leak. The paper claims that fusing the text into visual features, then passing the fused features through a temporal-spatial motion module and a false-positive removal step, lets the model segment blurry, non-rigid gas clouds better than existing image- or video-based methods. On the synthetic SimGas benchmark it reports a J&F score of 65.99 versus 58.76 for the best comparison; on the reorganized few-shot IGS-Few benchmark it reports 72.05. The stakes are practical: automated infrared surveillance that raises an alarm only when a leak is actually present.

What carries the argument

The load-bearing object is the Vision-Language Fusion (VLF) module, which multiplies visual features $f_v$ with text representations $f_t$ from the OWLv2 text encoder via a dot product $F_{vt}=(\phi(f_v\cdot f_t))^T$, yielding fused features at each scale. The Temporal-Spatial Module (TSM) then computes a normalized 4D correlation volume between consecutive frames to expose subtle motion and refines it through granular group-wise channel mixing; the decoder is a Feature Pyramid Network, and a morphological opening with kernel size 9 removes small false-positive blobs. The VLF supplies semantic guidance for unseen appearances, while the opening supplies clean outputs for frames with no leak.

What would settle it

Measure the size distribution of false-positive masks in the SimGas non-leak frames before the cleanup step. If any false-positive blob is wider than the nine-pixel filter, or any true leak contains components narrower than the filter, the cleanup cannot do what the paper asks of it; a direct check is to sweep the filter size from 3 to 15 and plot J&F against per-video leak size to see whether small plumes are erased.

Watch

Extended reading notes

Core claim

The paper's central claim is that a single architecture can handle both supervised and few-shot gas leak segmentation better than prior specialized models. JVLGS combines a Pyramid Vision Transformer vision encoder with the OWLv2 text encoder, a dot-product fusion that injects prompt semantics into every scale, a Temporal-Spatial Module that builds correlation volumes between neighboring frames and refines them with granular spatial mixing, and a Feature Pyramid Network decoder. A final morphological opening with a nine-pixel kernel erases small noise masks so that non-leak frames come out entirely black. The reported J&F scores are 65.99 on SimGas and 72.05 on IGS-Few, which the paper says outperform all compared methods, and the ablation attributes the SimGas gain mainly to the post-processing step while attributing generalization to unseen scenarios mainly to the text prompts.

Load-bearing premise

The weakest premise is that a single cleanup filter, tuned on the test sets, can delete every false-positive blob while preserving every real leak; the ablation shows this filter, not the vision-language fusion, produces most of the SimGas gain, so if false alarms or leaks come in different sizes the reported edge disappears.

Editorial extensions

If this is right

  • If correct, text-prompt conditioning becomes a viable route for segmenting semi-transparent, non-rigid targets whose shape and texture cues are unreliable.
  • The same model can work with full supervision and with only about 30 training frames per video, matching the few-shot reality of surveillance-site deployment.
  • Producing entirely black masks on no-leak frames means JVLGS could support alarm systems that are not triggered by clouds, steam, or camera noise.
  • The reported scores place JVLGS ahead of the specialized gas-leak models it compares with on both benchmarks, making it a new reference point for the task.

Reading between the lines

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

  • The ablation suggests the text prompts matter mostly on SimGas, where scenarios are unseen; on IGS-Few the prompts add almost nothing, so their role is probably generalization to unfamiliar appearances rather than tighter segmentation of known ones.
  • The dominant SimGas gain comes from the cleanup step (J&F jumps from 41.78 to 65.99 when it is added), whereas the vision-language fusion adds about 2.76 points, so the publicized improvement depends heavily on the fixed cleanup kernel.
  • A testable extension is to replace the fixed opening with an adaptive or learned false-positive filter and check whether the SimGas gain survives across different leak plume sizes and camera distances.
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

3 major / 5 minor

Summary. The paper proposes JVLGS, a video-based gas leak segmentation framework that combines a PVTv2 vision encoder, a vision-language fusion (VLF) module built on the OWLv2 text encoder with four hand-crafted prompts, a temporal-spatial module (TSM), an FPN decoder, and a morphological-opening post-processing step. Experiments on the SimGas and IGS-Few datasets report state-of-the-art J&F scores of 65.99 and 72.05, respectively, under supervised and few-shot settings, and the authors claim that the vision-language fusion, temporal-spatial module, and false-positive post-processing jointly drive the gains.

Significance. If the reported results were obtained under a sound evaluation protocol, the work would provide a useful engineering contribution to infrared gas leak segmentation, with the strengths of released code, a k-fold cross-validation protocol on SimGas, and a standardized metric computation across baselines. However, the manuscript's central claims are weakened by two internal problems: the ablation table contradicts the claimed benefit of the vision-language fusion module, and the headline performance relies on post-processing and prompt hyperparameters that appear to be selected on the test sets themselves. As presented, the evidence does not establish that the proposed architectural components—rather than test-set-tuned post-processing—are responsible for the reported state-of-the-art results.

major comments (3)
  1. [§4.4, Table 4] Rows 4 and 5 of Table 4 show that adding the VLF module to the TSM+FPN baseline without post-processing decreases SimGas J&F from 63.23 to 41.78; with the NCD decoder (Rows 1 vs 2) it decreases from 41.94 to 41.55. The text's 'Coordination Between Modules' explanation is post hoc and does not supply a controlled experiment that isolates the VLF contribution while keeping all other components fixed. In the only row pair where VLF appears beneficial (Row 4 vs Row 6), the comparison is confounded by the simultaneous addition of post-processing. On IGS-Few, the VLF module contributes essentially nothing (72.03 vs 72.05, Rows 4 vs 6). Thus the paper's core architectural novelty lacks direct experimental support.
  2. [§4.4, Figure 6 and Appendix B, Table 6] The post-processing kernel size 9 is selected by maximizing J&F on the test sets themselves (Figure 6), and the four text prompts are chosen by their per-video test-set contributions (Appendix B, Table 6). This is fitting hyperparameters to the test data, so the reported SimGas score of 65.99 is an optimistic in-sample estimate rather than a valid measure of expected performance on unseen videos. The manuscript should select the kernel size and prompt set on a validation split (or fix them a priori) and report the sensitivity of the headline numbers to these choices, including results without the post-processing step. The size prior claimed in Appendix D ('the size of non-leak masks is generally smaller than leak masks') is also not validated on unseen data and could erase small true leaks in deployment.
  3. [§4.3.2 and §4.4 (Table 1 vs Table 4)] The SimGas state-of-the-art claim rests entirely on the 24.21-point improvement from morphological opening (Row 5: 41.78 to Row 6: 65.99 in Table 4). Without this post-processing, JVLGS at 41.78 is below LangGas (58.76) and only marginally above FGSTP (41.94). Because the post-processing is a generic morphological filter whose kernel is tuned on test data, the paper has not established that the proposed vision-language architecture outperforms prior methods; moreover, the paper does not report whether applying the same post-processing to the baseline methods' outputs would close the gap. The claimed superiority is therefore not attributable to the proposed modules without additional experiments.
minor comments (5)
  1. [§3.5] The loss functions wBCE and wIoU are only referenced to [42]; providing the exact weighting scheme in the text or appendix would improve reproducibility.
  2. [§4.1] The IGS-Few re-split is described only verbally; the exact train/test video-frame assignment should be released along with the code to make comparisons reproducible.
  3. [Figure 6] The x-axis label 'kernel size' is ambiguous; please specify whether it is the side length or the radius of the structuring element.
  4. [Appendix D] The asserted size prior ('non-leak masks are generally smaller than leak masks') should be quantified on the datasets (for example, by showing the distribution of connected-component sizes) to justify the chosen kernel size.
  5. [Appendix F, Table 7] The layout of Table 7 appears corrupted in the manuscript, with fold weights and video names interleaved; please fix the formatting.

Circularity Check

2 steps flagged · score 6.0 of 10

SimGas SOTA is partially manufactured by test-set-tuned post-processing and test-set-selected prompts; the architecture itself remains independent.

  1. fitted input called prediction [Section 4.4, Figure 6; Table 4 Row 5 vs Row 6]
    "Post-Processing Optimal Kernel Size. We also explore the optimal kernel size for the post-processing, with results shown in Figure 6. Both excessively small and large kernels degrade performance: small kernels fail to remove false positives, while large kernels tend to over-erode valid mask regions, making them unrecoverable during dilation. Our experiments identify a kernel size of 9 as delivering the best performance across both datasets, and we adopt this value as a hyperparameter in our model."

    Figure 6 plots final J&F on the two test sets as a function of kernel size, and kernel 9 is selected as the argmax of that curve. Table 4 then presents Row 6 (with post-processing) as the model's headline result, attributing the decisive SimGas gain from 41.78 (Row 5) to 65.99 (Row 6) to this component. Consequently, the reported SimGas superiority is not an independent prediction of a fixed JVLGS pipeline: it is the selected maximum of a test-set fit over the morphological-opening parameter. Appendix D's size prior, that non-leak false-positive masks are 'generally smaller' than true leak masks, is not validated on held-out data, so the post-processing step that carries most of the SimGas margin is tuned to the benchmark rather than derived from first principles.

  2. fitted input called prediction [Appendix B, Table 6; Section 3.2 prompt-set construction]
    "We define prompt contribution as the average increase in J&F score relative to the non-prompt baseline, calculated across all videos. ... Prompts ... "White Steam" 15.03 21 ... All (JVLGS) 26.35 28. ... we use a total of four prompt sequences during training. The effectiveness of each sequence is detailed in Appendix B."

    The four prompts used in the main results were selected because they produce large positive J&F contributions on the test videos of SimGas, with 'White Steam' contributing +15.03 J&F and improving 21 videos. The reported JVLGS performance is then obtained using this same test-selected prompt set, so the claim that the language prompts drive performance on unseen scenarios is partly an artifact of choosing the prompts on the benchmark itself. Appendix B even concedes that prompts add almost nothing on IGS-Few (72.03 to 72.05), concentrating the entire prompt benefit on the dataset where the prompt list was optimized.

full rationale

The architecture itself is a self-contained engineering contribution: the VLF fusion, TCV, GSA, FPN decoder, and loss are defined by equations (1)-(9) and do not reduce to a cited uniqueness theorem or to one another. No load-bearing self-citation was found; references such as [7], [29], and [57] supply baselines or architectural inspirations rather than forced conclusions. The circularity is confined to the empirical evaluation: two inputs to the final method, the morphological-opening kernel size and the text-prompt set, are tuned directly against the test sets whose headline scores are reported. Because the largest reported SimGas gain comes from the test-tuned opening (Table 4) and the prompt contribution is likewise measured and selected on test videos (Appendix B), the central claim of outperforming all methods on SimGas is partially constructed by benchmark fitting rather than by an independent predictive test. The IGS-Few comparison is less affected, and the vision-language architecture remains an independent design contribution, so a moderate score of 6 is appropriate rather than a higher one.

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

The central claim rests on the transferred usefulness of OWLv2 text representations, the transferability of COD10K-pretrained PVTv2, and the test-set-tuned opening kernel. The free parameters are the kernel size and the hand-written prompt set, both chosen using the evaluation data.

free parameters (3)
  • Post-processing kernel size = 9
    Chosen as the value maximizing J&F on both SimGas and IGS-Few test sets (Figure 6, Section 4.4). This is tuning on the evaluation data.
  • Text prompt set = ['White Steam', 'Floating Steam', 'Billowing Smoke', 'Blowing Smoke']
    Selected based on per-video J&F contribution on SimGas test videos (Appendix B, Table 6). Prompt contributions were measured on the test set and the best combination was adopted.
  • Number of prompt sequences = 4
    Chosen with the prompt set; no separate validation.
assumptions (4)
  • domain assumption The OWLv2 text encoder, repurposed for text-only representation, yields useful semantic guidance for gas leaks.
    Section 3.2: text representations are extracted from a model trained for open-vocabulary object detection, transferred to gas segmentation without task-specific text supervision.
  • domain assumption PVTv2 encoder pretrained on COD10K transfers to gas leak segmentation.
    Appendix A: The authors use COD10K-pretrained PVTv2 because training from scratch yields 26.53% lower J&F (Table 5). Transferability is assumed.
  • domain assumption The temporal correlation volume (Eq. 4) captures subtle gas motion cues.
    Borrowed from SLT-Net [7]; no analysis of failure modes for blurry gas clouds is given beyond qualitative examples.
  • ad hoc to paper Morphological opening with a 9x9 kernel removes false-positive masks without removing true leak regions.
    This is the load-bearing post-processing assumption, validated only on the two test datasets where the kernel was tuned (Section 4.4).

how reviews work

0 comments
Cite this review

Pith. "Pith review of JVLGS: Joint Vision-Language Gas Leak Segmentation." pith.science (2026). https://pith.science/paper/YTMASGFN

@misc{pith2026250819485,
  author       = {Pith},
  title        = {Pith review of: JVLGS: Joint Vision-Language Gas Leak Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YTMASGFN}},
  note         = {Machine review of arXiv:2508.19485}
}
read the original abstract

Gas leaks pose serious threats to human health and contribute significantly to atmospheric pollution, drawing increasing public concern. However, the lack of effective detection methods hampers timely and accurate identification of gas leaks. While some vision-based techniques leverage infrared videos for leak detection, the blurry and non-rigid nature of gas clouds often limits their effectiveness. To address these challenges, we propose a novel framework called Joint Vision-Language Gas leak Segmentation (JVLGS), which integrates the complementary strengths of visual and textual modalities to enhance gas leak representation and segmentation. Recognizing that gas leaks are sporadic and many video frames may contain no leak at all, our method incorporates a post-processing step to reduce false positives caused by noise and non-target objects, an issue that affects many existing approaches. Extensive experiments conducted across diverse scenarios show that JVLGS significantly outperforms state-of-the-art gas leak segmentation methods. We evaluate our model under both supervised and few-shot learning settings, and it consistently achieves strong performance in both, whereas competing methods tend to perform well in only one setting or poorly in both. Code available at: https://github.com/GeekEagle/JVLGS

Figures

Figures reproduced from arXiv: 2508.19485 by the authors.

Figure 1
Figure 1. Framework of Joint Vision-Language Gas leak Segmentation (JVLGS). The inputs are video clips and text prompts of the target [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Architecture of the Vision-Language Fusion (VLF) mod [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Framework of Temporal Spatial Module (TSM), which is constituted by two parts, Temporal Correlation Volume (TCV) and [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Visualized results on SimGas. Our model achieves the highest accuracy and generates whole-black masks for non-leak cases. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Visualized results on IGS-Few. Our model distinguishes the leak better than other models in few-shot learning. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: The relationship between [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: shows that the most influential prompt is “White Steam”, which significantly reduces the number of videos whose J &F scores are in the 0–20 range and shifts them into higher performance categories. Prompts like “Floating [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]
Figure 8
Figure 8. Figure 8: J &F value distributions of each prompt on IGS-Few dataset videos. Prompts Avg. J &F Contribution No. of Improved Videos No Prompt 0 0 “White Steam” 15.03 21 “Floating Steam” 10.26 12 “Billowing Smoke” 1.14 12 “Blowing Smoke” 8.17 18 All (JVLGS) 26.35 28 [PITH_FULL_IM…
Figure 9
Figure 9. Figure 9: Visualized contribution of text prompt. 14 [PITH_FULL_IMAGE:figures/full_fig_p014_9.png]
Figure 10
Figure 10. Figure 10: Visualized Results on SimGas dataset. Our model predicts more accurately than any other baseline model. And ours can give a [PITH_FULL_IMAGE:figures/full_fig_p015_10.png]
Figure 11
Figure 11. Figure 11: Visualized Results on IGS-Few dataset. Our model distinguishes the leak and background in various scenarios better than other [PITH_FULL_IMAGE:figures/full_fig_p016_11.png]
Figure 12
Figure 12. Figure 12: Results of false positive removal. E. Performance-Efficiency Analysis [PITH_FULL_IMAGE:figures/full_fig_p017_12.png]
Figure 13
Figure 13. Figure 13: Performance-Efficiency Diagram. average across all folds. The weight for each fold is de￾termined by the total number of frames in its correspond￾ing video group. Since the number of frames varies across videos and each fold consists of several grouped videos, the tot…

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

60 extracted references · 50 canonical work pages

  1. [1]

    Recent advances in pipeline monitoring and oil leak- age detection technologies: Principles and approaches

    Mutiu Adesina Adegboye, Wai-Keung Fung, and Aditya Karnik. Recent advances in pipeline monitoring and oil leak- age detection technologies: Principles and approaches. Sen- sors, 19(11), 2019. 1

    work page 2019

  2. [2]

    Leak detec- tion and repair compliance assistance guidance best practices guide - ldarguide.pdf, 2014

    United States Environmental Protection Agency. Leak detec- tion and repair compliance assistance guidance best practices guide - ldarguide.pdf, 2014. 2

    work page 2014

  3. [3]

    Enis C ¸ etin

    Diaa Badawi, Hongyi Pan, Sinan Cem Cetin, and A. Enis C ¸ etin. Computationally efficient spatio-temporal dy- namic texture recognition for volatile organic compound (voc) leakage detection in industrial plants. IEEE Journal of Selected Topics in Signal Processing, 14(4):676–687, 2020. 1

    work page 2020

  4. [4]

    Xmem++: Production-level video segmentation from few annotated frames

    Maksym Bekuzarov, Ariana Bermudez, Joon-Young Lee, and Hao Li. Xmem++: Production-level video segmentation from few annotated frames. InProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 635–644, October 2023. 1, 5

    work page 2023

  5. [5]

    Stcnet: spa- tiotemporal cross network for industrial smoke detection

    Yichao Cao, Qingfei Tang, and Xiaobo Lu. Stcnet: spa- tiotemporal cross network for industrial smoke detection. Multimedia Tools and Applications , 81(7):10261–10277,

    work page

  6. [6]

    Putting the object back into video object segmentation

    Ho Kei Cheng, Seoung Wug Oh, Brian Price, Joon-Young Lee, and Alexander Schwing. Putting the object back into video object segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 3151–3161, June 2024. 1, 6

    work page 2024

  7. [7]

    Im- plicit motion handling for video camouflaged object detec- tion

    Xuelian Cheng, Huan Xiong, Deng-Ping Fan, Yiran Zhong, Mehrtash Harandi, Tom Drummond, and Zongyuan Ge. Im- plicit motion handling for video camouflaged object detec- tion. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR), pages 13864– 13873, June 2022. 1, 2, 4, 5, 6, 19

    work page 2022

  8. [8]

    Maskclip: Masked self-distillation advances contrastive language-image pretraining, 2023

    Xiaoyi Dong, Jianmin Bao, Yinglin Zheng, Ting Zhang, Dongdong Chen, Hao Yang, Ming Zeng, Weiming Zhang, Lu Yuan, Dong Chen, Fang Wen, and Nenghai Yu. Maskclip: Masked self-distillation advances contrastive language-image pretraining, 2023. 2

    2023

Show all 60 references
  1. [9]

    An image is worth 16x16 words: Trans- formers for image recognition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, et al. An image is worth 16x16 words: Trans- formers for image recognition at scale. arXiv preprint a...

    2010 arXiv

  2. [10]

    Concealed object detection

    Deng-Ping Fan, Ge-Peng Ji, Ming-Ming Cheng, and Ling Shao. Concealed object detection. IEEE Transactions on Pattern Analysis and Machine Intelligence , 44(10):6024– 6042, 2021. 1, 5, 6, 12, 19

    2021

  3. [11]

    Advised: Advanced video smoke detection for real-time measurements in antifire indoor and outdoor systems

    Alessio Gagliardi and Sergio Saponara. Advised: Advanced video smoke detection for real-time measurements in antifire indoor and outdoor systems. Energies, 13(8):2098, 2020. 2

    2020

  4. [12]

    Texture segmentation by multiscale aggregation of filter responses and shape ele- ments

    Galun, Sharon, Basri, and Brandt. Texture segmentation by multiscale aggregation of filter responses and shape ele- ments. In Proceedings Ninth IEEE international conference on computer vision, pages 716–723. IEEE, 2003. 2

    2003

  5. [13]

    Langgas: Introducing language in selective zero-shot background subtraction for semi-transparent gas leak detection with a new dataset.arXiv preprint arXiv:2503.02910, 2025

    Wenqi Guo, Yiyang Du, and Shan Du. Langgas: Introducing language in selective zero-shot background subtraction for semi-transparent gas leak detection with a new dataset.arXiv preprint arXiv:2503.02910, 2025. 5, 6, 7, 19

    2025 arXiv

  6. [14]

    Strategic preys make acute predators: Enhancing camouflaged object detectors by generating camouflaged objects

    Chunming He, Kai Li, Yachao Zhang, Yulun Zhang, Zhen- hua Guo, Xiu Li, Martin Danelljan, and Fisher Yu. Strategic preys make acute predators: Enhancing camouflaged object detectors by generating camouflaged objects. arXiv preprint arXiv:2308.03166, 2023. 2

    2023 arXiv

  7. [15]

    Decoupling static and hier- archical motion perception for referring video segmentation

    Shuting He and Henghui Ding. Decoupling static and hier- archical motion perception for referring video segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 13332–13341, 2024. 3, 5

    2024

  8. [16]

    Implicit-explicit motion learning for video cam- ouflaged object detection

    Wenjun Hui, Zhenfeng Zhu, Guanghua Gu, Meiqin Liu, and Yao Zhao. Implicit-explicit motion learning for video cam- ouflaged object detection. IEEE Transactions on Multime- dia, 26:7188–7196, 2024. 2

    2024

  9. [17]

    Endow sam with keen eyes: Temporal-spatial prompt learn- ing for video camouflaged object detection

    Wenjun Hui, Zhenfeng Zhu, Shuai Zheng, and Yao Zhao. Endow sam with keen eyes: Temporal-spatial prompt learn- ing for video camouflaged object detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, pages 19058–19067, 2024. 2

    2024

  10. [18]

    Fast camouflaged object detection via edge-based reversible re- calibration network

    Ge-Peng Ji, Lei Zhu, Mingchen Zhuge, and Keren Fu. Fast camouflaged object detection via edge-based reversible re- calibration network. Pattern Recognition, 123:108414, 2022. 2

    2022

  11. [19]

    Gas leakage detection using spatial and temporal neural network model

    Alibek Kopbayev, Faisal Khan, Ming Yang, and Syeda Zohra Halim. Gas leakage detection using spatial and temporal neural network model. Process Safety and Environmental Protection, 160:968–975, 2022. 2

    2022

  12. [20]

    Betrayed by motion: Camouflaged object discovery via motion segmentation

    Hala Lamdouar, Charig Yang, Weidi Xie, and Andrew Zis- serman. Betrayed by motion: Camouflaged object discovery via motion segmentation. In Proceedings of the Asian con- ference on computer vision, 2020. 2

    2020

  13. [21]

    Visualbert: A simple and perfor- mant baseline for vision and language

    Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, and Kai-Wei Chang. Visualbert: A simple and perfor- mant baseline for vision and language. arXiv preprint arXiv:1908.03557, 2019. 2

    1908 arXiv

  14. [22]

    Grounded language-image pre-training

    Liunian Harold Li, Pengchuan Zhang, Haotian Zhang, Jian- wei Yang, Chunyuan Li, Yiwu Zhong, Lijuan Wang, Lu Yuan, Lei Zhang, Jenq-Neng Hwang, et al. Grounded language-image pre-training. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, page...

    2022

  15. [23]

    Feature pyramid networks for object detection, 2017

    Tsung-Yi Lin, Piotr Doll ´ar, Ross Girshick, Kaiming He, Bharath Hariharan, and Serge Belongie. Feature pyramid networks for object detection, 2017. 3

    2017

  16. [24]

    An effective low-contrast SF6 gas leakage detection method for infrared imaging

    Quan Lu, Qin Li, Likun Hu, and Lifeng Huang. An effective low-contrast SF6 gas leakage detection method for infrared imaging. IEEE Transactions on Instrumentation and Mea- surement, 70:1–9, 2021. 1

    2021

  17. [25]

    Image segmenta- tion using text and image prompts

    Timo L ¨uddecke and Alexander Ecker. Image segmenta- tion using text and image prompts. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 7086–7096, 2022. 2 9

    2022

  18. [26]

    Scaling open-vocabulary object detection

    Matthias Minderer, Alexey Gritsenko, and Neil Houlsby. Scaling open-vocabulary object detection. Advances in Neu- ral Information Processing Systems, 36:72983–73007, 2023. 2, 3

    2023

  19. [27]

    Simple open-vocabulary object detection

    Matthias Minderer, Alexey Gritsenko, Austin Stone, Maxim Neumann, Dirk Weissenborn, Alexey Dosovitskiy, Aravindh Mahendran, Anurag Arnab, Mostafa Dehghani, Zhuoran Shen, et al. Simple open-vocabulary object detection. In European conference on computer vision , pages 728–755. ...

    2022

  20. [28]

    Actionvos: Actions as prompts for video object segmentation

    Liangyang Ouyang, Ruicong Liu, Yifei Huang, Ryosuke Fu- ruta, and Yoichi Sato. Actionvos: Actions as prompts for video object segmentation. pages 216–235, 2024. 3

    2024

  21. [29]

    Zoomnext: A unified collaborative pyra- mid network for camouflaged object detection

    Youwei Pang, Xiaoqi Zhao, Tian-Zhu Xiang, Lihe Zhang, and Huchuan Lu. Zoomnext: A unified collaborative pyra- mid network for camouflaged object detection. IEEE Trans- actions on Pattern Analysis and Machine Intelligence, 2024. 2, 4, 6, 19

    2024

  22. [30]

    A benchmark dataset and evaluation methodology for video object segmentation

    Federico Perazzi, Jordi Pont-Tuset, Brian McWilliams, Luc Van Gool, Markus Gross, and Alexander Sorkine-Hornung. A benchmark dataset and evaluation methodology for video object segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition, page...

  23. [31]

    Learning transferable visual models from natural language supervi- sion

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervi- sion. In International conference on machine learning, ...

    2021

  24. [32]

    Zero-shot text-to-image generation

    Aditya Ramesh, Mikhail Pavlov, Gabriel Goh, Scott Gray, Chelsea V oss, Alec Radford, Mark Chen, and Ilya Sutskever. Zero-shot text-to-image generation. In International confer- ence on machine learning, pages 8821–8831. Pmlr, 2021. 2, 5

    2021

  25. [33]

    Are optical gas imaging technologies effective for methane leak detection? Environmental science & technology , 51(1):718–724, 2017

    Arvind P Ravikumar, Jingfan Wang, and Adam R Brandt. Are optical gas imaging technologies effective for methane leak detection? Environmental science & technology , 51(1):718–724, 2017. 1

    2017

  26. [34]

    Opening and closing

    Pierre Soille and Pierre Soille. Opening and closing. Mor- phological image analysis: Principles and applications , pages 105–137, 2004. 5

    2004

  27. [35]

    Detection of regions of interest and camouflage breaking by direct convexity estima- tion

    Ariel Tankus and Yehezkel Yeshurun. Detection of regions of interest and camouflage breaking by direct convexity estima- tion. In Proceedings 1998 IEEE workshop on visual surveil- lance, pages 42–48. IEEE, 1998. 2

    1998

  28. [36]

    Yolov7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors

    Chien-Yao Wang, Alexey Bochkovskiy, and Hong- Yuan Mark Liao. Yolov7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages 7464–7475, June 2023. 1

    2023

  29. [37]

    Videogasnet: Deep learning for natural gas methane leak classification using an infrared camera

    Jingfan Wang, Jingwei Ji, Arvind P Ravikumar, Silvio Savarese, and Adam R Brandt. Videogasnet: Deep learning for natural gas methane leak classification using an infrared camera. Energy, 238:121516, 2022. 2, 5

    2022

  30. [38]

    Tchapmi, Arvind P

    Jingfan Wang, Lyne P. Tchapmi, Arvind P. Ravikumar, Mike McGuire, Clay S. Bell, Daniel Zimmerle, Silvio Savarese, and Adam R. Brandt. Machine vision for natural gas methane emissions detection using an infrared camera. Ap- plied Energy, 257:113998, 2020. 1

    2020

  31. [39]

    Deep learning-based object detection for vis- ible dust and prevention measures on construction sites

    Mingpu Wang, Gang Yao, Yang Yang, Yujia Sun, Meng Yan, and Rui Deng. Deep learning-based object detection for vis- ible dust and prevention measures on construction sites. De- velopments in the Built Environment, 16:100245, 2023. 1

    2023

  32. [40]

    Pvt v2: Improved baselines with pyramid vision transformer

    Wenhai Wang, Enze Xie, Xiang Li, Deng-Ping Fan, Kaitao Song, Ding Liang, Tong Lu, Ping Luo, and Ling Shao. Pvt v2: Improved baselines with pyramid vision transformer. Computational visual media, 8(3):415–424, 2022. 3, 12

    2022

  33. [41]

    Fire smoke detection based on texture fea- tures and optical flow vector of contour

    Yakun Wang, Aiguo Wu, Jie Zhang, Meng Zhao, Wenshuai Li, and Na Dong. Fire smoke detection based on texture fea- tures and optical flow vector of contour. In 2016 12th World Congress on Intelligent Control and Automation (WCICA) , pages 2879–2883, 2016. 1

    2016

  34. [42]

    F3net: fusion, feedback and focus for salient object detection

    Jun Wei, Shuhui Wang, and Qingming Huang. F3net: fusion, feedback and focus for salient object detection. In Proceed- ings of the AAAI conference on artificial intelligence , vol- ume 34, pages 12321–12328, 2020. 5

    2020

  35. [43]

    A survey of camouflaged object detection and be- yond

    Fengyang Xiao, Sujie Hu, Yuqi Shen, Chengyu Fang, Jinfa Huang, Chunming He, Longxiang Tang, Ziyun Yang, and Xiu Li. A survey of camouflaged object detection and be- yond. arXiv preprint arXiv:2408.14562, 2024. 1

    2024 arXiv

  36. [44]

    Segmenting moving objects via an object-centric layered representation

    Junyu Xie, Weidi Xie, and Andrew Zisserman. Segmenting moving objects via an object-centric layered representation. In Proceedings of the 36th International Conference on Neu- ral Information Processing Systems , NIPS ’22, Red Hook, NY , USA, 2022. Curran Associates Inc. 2

    2022

  37. [45]

    Flame and smoke detection in substation based on wavelet analysis and convolution neural network

    Zhang Xu, Wang Wanguo, Li Xinrui, Liu Bin, and Tian Yuan. Flame and smoke detection in substation based on wavelet analysis and convolution neural network. In Pro- ceedings of the 2019 3rd International Conference on Inno- vation in Artificial Intelligence, pages 248–252, 2019. 2

    2019

  38. [46]

    Visa: Reasoning video object segmentation via large language models

    Cilin Yan, Haochen Wang, Shilin Yan, Xiaolong Jiang, Yao Hu, Guoliang Kang, Weidi Xie, and Efstratios Gavves. Visa: Reasoning video object segmentation via large language models. arXiv preprint arXiv:2407.11325, 2024. 3

    2024 arXiv

  39. [47]

    Mirror- net: Bio-inspired camouflaged object segmentation

    Jinnan Yan, Trung-Nghia Le, Khanh-Duy Nguyen, Minh- Triet Tran, Thanh-Toan Do, and Tam V Nguyen. Mirror- net: Bio-inspired camouflaged object segmentation. IEEE Access, 9:43290–43300, 2021. 2

    2021

  40. [48]

    Semi-supervised video salient object detection using pseudo-labels

    Pengxiang Yan, Guanbin Li, Yuan Xie, Zhen Li, Chuan Wang, Tianshui Chen, and Liang Lin. Semi-supervised video salient object detection using pseudo-labels. In Proceedings of the IEEE International Conference on Computer Vision , pages 7284–7293, 2019. 1

    2019

  41. [49]

    Self-supervised video object segmentation by motion grouping

    Charig Yang, Hala Lamdouar, Erika Lu, Andrew Zisserman, and Weidi Xie. Self-supervised video object segmentation by motion grouping. In Proceedings of the IEEE/CVF In- ternational Conference on Computer Vision (ICCV) , pages 7177–7188, October 2021. 1

    2021

  42. [50]

    Dynamic texture based smoke detection using surfacelet transform and hmt model

    Wei Ye, Jianhui Zhao, Song Wang, Yong Wang, Dengyi Zhang, and Zhiyong Yuan. Dynamic texture based smoke detection using surfacelet transform and hmt model. Fire Safety Journal, 73:91–101, 2015. 2 10

    2015

  43. [51]

    Optical gas detection: key technologies and applica- tions review

    Shen Ying, Shao Kunming, Wu Jing, Huang Feng, and Guo Yuze. Optical gas detection: key technologies and applica- tions review. Opto-Electronic Engineering, 47(4):190280–1,

  44. [52]

    A lightweight network based on local-global feature fusion for real-time industrial invisible gas detection with infrared ther- mography

    Huan Yu, Jin Wang, Zhan Wang, Jingru Yang, Kaixiang Huang, Guodong Lu, Fengtao Deng, and Yang Zhou. A lightweight network based on local-global feature fusion for real-time industrial invisible gas detection with infrared ther- mography. Applied Soft Computing, 152:111138, 202...

    2024

  45. [53]

    Leak detection and localization techniques in oil and gas pipeline: A bibliometric and sys- tematic review

    Jie Yuan, Wenjing Mao, Chun Hu, Jianfeng Zheng, Dezhi Zheng, and Yunbo Yang. Leak detection and localization techniques in oil and gas pipeline: A bibliometric and sys- tematic review. Engineering Failure Analysis, 146:107060,

  46. [54]

    Vision-language models for vision tasks: A survey

    Jingyi Zhang, Jiaxing Huang, Sheng Jin, and Shijian Lu. Vision-language models for vision tasks: A survey. IEEE Trans. Pattern Anal. Mach. Intell. , 46(8):5625–5644, Aug

  47. [55]

    Explicit motion handling and interactive prompting for video camouflaged object detection

    Xin Zhang, Tao Xiao, Ge-Peng Ji, Xuan Wu, Keren Fu, and Qijun Zhao. Explicit motion handling and interactive prompting for video camouflaged object detection. Trans. Img. Proc., 34:2853–2866, May 2025. 2

    2025

  48. [56]

    A survey of large language models

    Wayne Xin Zhao, Kun Zhou, Junyi Li, Tianyi Tang, Xiaolei Wang, Yupeng Hou, Yingqian Min, Beichen Zhang, Junjie Zhang, Zican Dong, Yifan Du, Chen Yang, Yushuo Chen, Zhipeng Chen, Jinhao Jiang, Ruiyang Ren, Yifan Li, Xinyu Tang, Zikang Liu, Peiyu Liu, Jian-Yun Nie, and Ji-Rong W...

    2023 arXiv

  49. [57]

    Fine-grained spatial-temporal perception for gas leak segmentation

    Xinlong Zhao and Shan Du. Fine-grained spatial-temporal perception for gas leak segmentation. arXiv: 2505.00295 ,

  50. [58]

    Rmem: Restricted memory banks improve video object segmenta- tion

    Junbao Zhou, Ziqi Pang, and Yu-Xiong Wang. Rmem: Restricted memory banks improve video object segmenta- tion. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR), pages 18602– 18611, June 2024. 1

    2024

  51. [59]

    Clip-vis: Adapting clip for open-vocabulary video instance segmentation

    Wenqi Zhu, Jiale Cao, Jin Xie, Shuangming Yang, and Yan- wei Pang. Clip-vis: Adapting clip for open-vocabulary video instance segmentation. IEEE Trans. Cir. and Sys. for Video Technol, 35(2):1098–1110, Feb 2025. 3, 5

    2025

  52. [60]

    White Steam

    Daniel Zimmerle, Timothy Vaughn, Clay Bell, Kristine Ben- nett, Parik Deshmukh, and Eben Thoma. Detection limits of optical gas imaging for natural gas leak detection in realistic controlled conditions. Environmental Science & Technology, 54(18):11506–11514, 2020. PMID: 327865...

    2020

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.