Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

DIS-Mine: Instance Segmentation for Disaster-Awareness in Poor-Light Condition in Underground Mines

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 DIS-Mine fuses SAM and Mask R-CNN to segment disaster-affected underground mine scenes in poor light, reporting 86.0% F1 and 72.0% mIoU on DsLMF+ and 70.2%/60.5% on its new ImageMine dataset.

desk verdict The ImageMine dataset is a real asset, but the headline numbers are overclaimed and the evaluation on ImageMine is undermined by auto-annotated labels from the same SAM model that DIS-Mine uses. read the letter →

arxiv 2411.13544 v1 pith:VRMDK4UM submitted 2024-11-20 cs.CV

Mizanur Rahman Jewel , Mohamed Elmahallawy , Sanjay Madria , Samuel Frimpong This is my paper · ORCID
classification cs.CV
keywords instancesegmentationlow-lightimageenhancementundergroundminingsafetydisasterresponseSegmentAnythingModelMaskR-CNNdatasetalignment
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

DIS-Mine tries to establish that instance segmentation in near-dark underground mines can be made reliable enough for rescue use, despite the noise, color distortion, and low contrast that defeat standard models. The method couples a KinD-based brightness-enhancement network with two complementary segmenters—a fine-tuned Segment Anything Model and a Mask R-CNN—and fuses their masks through ORB feature matching, intersection, and morphological cleanup. The authors also introduce ImageMine, a real-world dataset of dark experimental-mine footage labeled for people, equipment, corridors, and surrounding structures. On DsLMF+ they report 86.0% F1 and 72.0% mIoU, beating SAM, Mask R-CNN, and Mask2Former, and on ImageMine they report 70.2% F1 and 60.5% mIoU. If the method works as claimed, first responders could receive instance-level maps of people and hazards in otherwise invisible mine conditions.

What carries the argument

The load-bearing mechanism is the mask-alignment fusion of two heterogeneous segmenters. A KinD (Retinex-based) network first brightens the dark image. A fine-tuned SAM takes the enhanced image plus bounding-box and class prompts generated by Mask R-CNN and produces one mask; Mask R-CNN, trained with a weighted-Dice-plus-focal-loss mask term, produces another mask. ORB feature matching aligns the two masks, their intersection is taken, and dilation/erosion plus a 4x4-grid rule (e.g., anything in the lower row is labeled road) yield the final instance mask. The idea is that the two models' errors are different enough that intersecting their aligned outputs removes false positives while the feature-matching step fixes spatial misalignment.

What would settle it

Take a hold-out set of dark mine images, label every mask by hand without any SAM assistance, then evaluate DIS-Mine and the baselines on that split; if DIS-Mine's F1/mIoU advantage over SAM and Mask R-CNN collapses, the reported superiority is an artifact of sharing the SAM label generator rather than a genuine property of the fusion.

Watch

Extended reading notes

Core claim

The paper's central claim is that its four-component pipeline—image brightness improvement, SAM-based instance segmentation with prompts from Mask R-CNN, Mask R-CNN with a weighted-Dice-plus-focal mask loss, and final mask alignment by ORB feature matching—outperforms each component alone and every single-model state-of-the-art baseline under poor-light conditions. The reported evidence is Table II: on DsLMF+ DIS-Mine reaches 86.0% F1/72.0% mIoU (SAM: 84.0/71.0; Mask R-CNN: 80.0/68.0; Mask2Former: 83.0/72.0); on ImageMine it reaches 70.2/60.5 (SAM: 68.7/60.0; Mask R-CNN: 65.0/56.0; Mask2Former: 67.2/58.0); and on LIS it reaches 63.2/47.0, above ISD's 61.7/49.8 on F1 but below on mIoU. The abstract's additional statement of 'at least 15x improvement and up to 80% higher precision in object detection' is not backed by a comparison table in the manuscript.

Load-bearing premise

The load-bearing premise is that the ImageMine ground-truth masks used for training and evaluation are accurate and independent of the SAM model embedded in DIS-Mine, even though Section III.C uses a fine-tuned SAM to auto-annotate the dataset and Section V.A trains Mask R-CNN on those auto-generated masks.

Editorial extensions

If this is right

  • Deployed on mine surveillance feeds, DIS-Mine could give rescue teams near-real-time instance masks of people, equipment, and corridors in darkness.
  • The automatic SAM-based annotation pipeline can label new images from similar underground environments, reducing the manual labeling burden from thousands of images to a few hundred.
  • Fusing two segmenters with alignment beats each segmenter alone on all three datasets reported, suggesting the fusion strategy, not just better features, drives the gains.
  • On the LIS benchmark DIS-Mine's F1 (63.2%) tops the specialized low-light segmenter ISD (61.7%), indicating the method transfers beyond mines.
  • The reported margins on DsLMF+ are small (1–6 points over SAM), so the practical benefit is a moderate, consistent improvement rather than a step change.

Reading between the lines

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

  • If the fusion advantage is real, a natural deployment is untethered rescue robots: DIS-Mine's pipeline could run on the enhanced image stream and broadcast instance masks of people and corridors to surface teams; the paper does not test this.
  • A direct test of the mask-alignment step would be to ablate the ORB intersection and keep either the SAM or Mask R-CNN branch; the paper reports no per-component ablation, so the contribution of alignment to the 70.2/60.5 scores is untested.
  • Because the auto-annotation pipeline uses SAM, an independent human-labeled test split would separate the method's true generalization from the label generator's familiarity; the paper does not provide one.
  • The stated future direction of fusing thermal or LiDAR modalities is a natural next step; in total darkness, active sensors would supply geometric cues that RGB enhancement cannot recover.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes DIS-Mine, an instance-segmentation pipeline for underground mine images under poor-light or near-dark conditions. The pipeline consists of four components: a KinD-based brightness-enhancement network, a fine-tuned Segment Anything Model (SAM) that takes Mask R-CNN predictions as prompts, a Mask R-CNN variant with weighted Dice plus focal mask loss, and a mask-alignment stage that uses ORB feature matching, morphological operations, and a rule-based 4x4 grid positional classifier. The authors also introduce ImageMine, a low-light underground mine dataset with a subset of manually annotated images, and describe an automatic SAM-based annotation pipeline for the rest. Experiments compare DIS-Mine against SAM, Mask R-CNN, Mask2Former, and ISD on ImageMine, LIS, and DsLMF+, reporting F1 and mIoU. The paper claims superior performance, including an 86.0% F1 score and 72.0% mIoU, with 'at least 15x improvement' over state-of-the-art methods.

Significance. If the ImageMine results are validated by an independent label protocol, the dataset itself would be a useful contribution to disaster-response vision in underground mines, and the proposed SAM-plus-Mask R-CNN fusion with low-light enhancement is a reasonable engineering direction. The paper also makes the dataset publicly available, which is a strength. However, the central quantitative claims are not currently supported by the reported experiments: the headline numbers apply only to the DsLMF+ benchmark, and the ImageMine evaluation appears to depend on labels generated by the same SAM model that the proposed method integrates. Because the comparison against baselines is the core of the paper, this evaluation gap is load-bearing.

major comments (4)
  1. [Abstract and Table II] The abstract states that DIS-Mine 'achieves a superior F1 score of 86.0% and mIoU of 72.0%, outperforming state-of-the-art instance segmentation methods, with at least 15x improvement and up to 80% higher precision in object detection.' This is not supported by the paper's own Table II. The 86.0% F1 and 72.0% mIoU are the DsLMF+ results; on ImageMine, the dataset introduced by the paper, DIS-Mine obtains 70.2% F1 and 60.5% mIoU. No table or figure anywhere reports a '15x improvement' or an '80% higher precision' figure, and object-detection precision is not an evaluated metric in Section V. Please restate the abstract using per-dataset results and remove claims that no experiment supports.
  2. [Section III.C and Section V.A] The ImageMine evaluation may be circular. Section III.C states that SAM was fine-tuned on 510 manually annotated images and then used to auto-annotate the remainder of ImageMine. Section V.A states that these SAM-generated image-mask pairs were used to train the Mask R-CNN component and that SAM was retrained on both manually and automatically annotated data. Since DIS-Mine explicitly fuses SAM with Mask R-CNN, its outputs share the same label distribution that produces the ground-truth masks, whereas the baselines are evaluated against masks produced by a generator they do not contain. The paper does not describe a separate human-labeled test split, a human/SAM agreement measure, or any exclusion of auto-annotated images from evaluation. Without an independent test set that is not used in SAM fine-tuning or Mask R-CNN training, the reported advantage over non-SAM baselines on ImageMine is not trustworthy. A revision must use a held-out manual test set and report agreement between SAM auto-annotations and human labels.
  3. [Section V.A and Section V.C] The experimental protocol is underspecified in ways that affect the interpretation of the reported margins. The paper does not state the total number of images in ImageMine, the number of auto-annotated images, or the train/validation/test split for any dataset. It also does not define the matching threshold used to compute instance-level F1 and mIoU, nor say whether metrics are averaged over multiple runs. Without these details, the differences between DIS-Mine and the second-best models on ImageMine (70.2% versus 68.7% F1; 60.5% versus 60.0% mIoU) cannot be assessed for significance or reproducibility. Please provide the full protocol, including dataset statistics and evaluation thresholds.
  4. [Section IV.D and Section V.C] No ablation study is reported, although the paper's claimed innovations are the four components of DIS-Mine. In particular, the rule-based 4x4 grid assignment in Section IV.D, which classifies objects as road/wall/roof according to their position in the image, is a strong inductive bias that could inflate performance on geometrically similar classes. The paper should add ablations that remove or deactivate (i) the brightness-improvement component, (ii) the SAM integration, (iii) the enhanced mask loss, and (iv) the positional grid rule, and report each variant's F1 and mIoU on ImageMine. This is necessary to determine which component is responsible for the reported gains.
minor comments (6)
  1. [Abstract and Introduction] The paper is inconsistent about the number of annotated classes: the Introduction says 'manually annotated a subset of images into five classes' and then later lists six categories (road, wall, roof, people, equipment, corridor). Please correct this inconsistency.
  2. [Section III.A] The phrase 'about 100-80 images were extracted' should presumably read '80-100 images'; please fix the range.
  3. [Section IV.A] Algorithm 1's caption says 'offers a explanation'; this should be 'offers an explanation'.
  4. [Section IV.D] The acronym ORB is expanded as 'oriented fAST and rotated brief'; the correct capitalization is 'Oriented FAST and Rotated BRIEF'.
  5. [Section V.C] The text says Mask2Former 'consistently ranks as the second-best model' after DIS-Mine, but in the LIS dataset row of Table II, ISD reports a higher F1 score (61.7%) than Mask2Former (62.0% is actually higher, so check: ISD 61.7%, Mask2Former 62.0%; the statement is true for F1, but for mIoU ISD is 49.8% versus Mask2Former 45.8%, so the sentence should be qualified by metric). Please make the ranking statement consistent with Table II.
  6. [Table III] For the People class, F1 and IoU are both reported as 72.6%; this coincidence is suspicious and should be checked, as F1 and IoU are different quantities.

Circularity Check

1 steps flagged · score 4.0 of 10

The ImageMine validation is partially circular: SAM fine-tuned on 510 manual images auto-annotates the rest of ImageMine, those masks train Mask R-CNN and are used as ground truth, while DIS-Mine itself fuses SAM; no independent human-labeled test split is reported. External LIS and DsLMF+ benchmarks give the central claim independent content.

  1. self definitional [Section III.C (Data annotation) and Section V.A (Training Procedure); evaluated in Table II (ImageMine rows)]
    "Initially, SAM was fine-tuned on a limited number of manually annotated datasets of 510 images... Once fine-tuned, it was applied for automatic annotation across the remainder of the dataset. ... Masks for each image were generated using the optimized SAM model, and these image-mask pairs were subsequently utilized to train our third component, the Mask R-CNN model. Additionally, the SAM model was retrained on both manually and automatically annotated datasets."

    The ImageMine ground-truth masks are not described as independent of DIS-Mine. The optimized SAM—the same model family that DIS-Mine integrates as its 'Instance Segmentation with SAM Integration' component—auto-annotates the bulk of ImageMine after being fine-tuned on 510 manual images; those auto-annotated masks are then used to train Mask R-CNN, and SAM is retrained on them. No separate human-labeled test split or human/SAM agreement measure is reported. Thus the Table II ImageMine F1/mIoU scores measure DIS-Mine against masks generated by one of its own components, while non-SAM baselines are scored against a SAM-specific label distribution. The margin over baselines is partly forced by label-generator overlap rather than by independent ground truth.

full rationale

The only circular step found is the ImageMine label-generation pipeline feeding the evaluation: the fine-tuned SAM both creates the auto-annotated ground truth and is a component of DIS-Mine, and the paper gives no evidence of a disjoint human-labeled test set. This makes the ImageMine comparison to non-SAM baselines partially self-referential. The rest of the derivation chain is independent: the KinD-based enhancement, Mask R-CNN loss modification, ORB alignment, and intersection/dilation/erosion fusion are standard components with no equation collapsing into the input. There are no load-bearing self-citations or uniqueness theorems. The abstract's 86.0/72.0 numbers are actually the DsLMF+ row of Table II (ImageMine is 70.2/60.5), and the 'at least 15x improvement' claim does not appear in any table; these are overclaiming/correctness concerns, not circularity. Because the headline result is also supported by external benchmarks, the circularity score is moderate rather than high.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claim rests on five unverified assumptions about the annotation loop, the synthetic degradation model, and hand-crafted alignment rules. The paper contributes no new physical entities; the main risk is that the SAM auto-annotation pipeline contaminates both training and evaluation labels.

free parameters (4)
  • Focal loss weighting alpha and focusing gamma = not reported
    Equation (5) defines the focal loss used in the mask branch; the values of alpha and gamma are never given in Table I or the text, yet they control segmentation behavior in dark regions.
  • Per-class weights w_c in weighted Dice loss = not reported
    Equation (4) introduces per-class weights w_c; no values are provided, and they can be tuned to favor specific classes, affecting the reported per-class F1 and IoU.
  • Morphological kernel sizes for dilation and erosion = not reported
    Algorithm 4 applies dilation and erosion to the combined mask but does not specify kernel sizes or iterations; these choices affect the final mask boundaries and therefore the metrics.
  • IoU matching threshold for F1 and mIoU in instance segmentation = not reported
    Instance-level F1 and mIoU depend on a mask matching threshold (commonly 0.5); the paper never states the threshold, so the reported numbers are not reproducible.
assumptions (5)
  • domain assumption KinD low-light enhancement trained on synthetic degradation pairs transfers to real underground mine images
    Section III.C creates low-light pairs by adding Gaussian noise, reducing brightness, and increasing contrast; there is no validation that this matches the real capture conditions in the mine.
  • domain assumption SAM auto-annotated masks are accurate enough to serve as ground truth for training Mask R-CNN
    Section III.C fine-tunes SAM on 510 manual images and uses it to label the rest; no agreement statistics between SAM masks and human labels are reported.
  • domain assumption ORB feature matching can align masks produced by two different segmentation architectures
    Section IV.D uses ORB on masks, but ORB was designed for natural images; its keypoints in dark, low-contrast mine images may be unreliable.
  • ad hoc to paper The 4x4 grid positional rule correctly assigns objects to classes such as road
    Section IV.D hand-codes a rule that any object in the lower row of the grid is road; this encodes a specific camera geometry and is not derived from data.
  • domain assumption Evaluation ground truth is independent of the SAM-based annotation pipeline
    The paper never states a human-labeled held-out test set; if evaluation uses SAM-generated masks, the comparison favors DIS-Mine because it embeds SAM.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DIS-Mine: Instance Segmentation for Disaster-Awareness in Poor-Light Condition in Underground Mines." pith.science (2026). https://pith.science/paper/VRMDK4UM

@misc{pith2026241113544,
  author       = {Pith},
  title        = {Pith review of: DIS-Mine: Instance Segmentation for Disaster-Awareness in Poor-Light Condition in Underground Mines},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VRMDK4UM}},
  note         = {Machine review of arXiv:2411.13544}
}
read the original abstract

Detecting disasters in underground mining, such as explosions and structural damage, has been a persistent challenge over the years. This problem is compounded for first responders, who often have no clear information about the extent or nature of the damage within the mine. The poor-light or even total darkness inside the mines makes rescue efforts incredibly difficult, leading to a tragic loss of life. In this paper, we propose a novel instance segmentation method called DIS-Mine, specifically designed to identify disaster-affected areas within underground mines under low-light or poor visibility conditions, aiding first responders in rescue efforts. DIS-Mine is capable of detecting objects in images, even in complete darkness, by addressing challenges such as high noise, color distortions, and reduced contrast. The key innovations of DIS-Mine are built upon four core components: i) Image brightness improvement, ii) Instance segmentation with SAM integration, iii) Mask R-CNN-based segmentation, and iv) Mask alignment with feature matching. On top of that, we have collected real-world images from an experimental underground mine, introducing a new dataset named ImageMine, specifically gathered in low-visibility conditions. This dataset serves to validate the performance of DIS-Mine in realistic, challenging environments. Our comprehensive experiments on the ImageMine dataset, as well as on various other datasets demonstrate that DIS-Mine achieves a superior F1 score of 86.0% and mIoU of 72.0%, outperforming state-of-the-art instance segmentation methods, with at least 15x improvement and up to 80% higher precision in object detection.

Figures

Figures reproduced from arXiv: 2411.13544 by the authors.

Figure 1
Figure 1. Samples from our underground mine dataset captured under [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Construction process for our ImageMine dataset. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. An overview of proposed DIS-Mine instance segmentation framework. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Illustration of the instance segmentation with SAM integration [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Evaluation of DIS-Mine against baselines on various metrics on the ImageMine dataset. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: DIS-Mine prediction results on samples from our collected ImageMine dataset. [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: DIS-Mine prediction results on samples from DsLMF+ dataset. [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Detecting Untargeted Attacks and Mitigating Unreliable Updates in Federated Learning for Underground Mining Operations

    cs.CR 2025-08 reject novelty 4.0 of 10

    MineDetect combines history-averaged gradient similarity, variance, and distance thresholds to flag sign-flip, noise, and unreliable clients in federated learning, but its headline accuracy gain is not borne out by th...

Reference graph

Works this paper leans on

28 extracted references · 21 canonical work pages · cited by 1 Pith paper

  1. [1]

    Current status of deep mining and disaster prevention in china,

    L. H. C. D. M. Debing, “Current status of deep mining and disaster prevention in china,” Coal Science and Technology , no. 1, 2016

    work page 2016

  2. [2]

    Residual coal exploitation and its impact on sustainable development of the coal industry in china,

    Y . Zhang et al. , “Residual coal exploitation and its impact on sustainable development of the coal industry in china,” Energy Policy, vol. 96, pp. 534–541, 2016

    work page 2016

  3. [3]

    Automation in the mining industry: Review of technology, systems, human factors, and political risk,

    W. P. Rogers et al. , “Automation in the mining industry: Review of technology, systems, human factors, and political risk,” Mining, metallurgy & exploration , vol. 36, pp. 607–631, 2019

    work page 2019

  4. [4]

    Tweedr: Mining twitter to inform disaster response

    Z. Ashktorab, C. Brown, M. Nandi, and A. Culotta, “Tweedr: Mining twitter to inform disaster response.” in ISCRAM, 2014, pp. 269–272

    work page 2014

  5. [5]

    Preparedness for peer first response to mining emergencies resulting in injuries: a cross-sectional study,

    S. Karlsson, B.-I. Saveman, M. Hultin, U. Björnstig, and L. Gyllencreutz, “Preparedness for peer first response to mining emergencies resulting in injuries: a cross-sectional study,” BMJ open, vol. 10, no. 11, 2020

    work page 2020

  6. [6]

    Towards an emergency preparedness for self- rescue from underground coal mines,

    M. Onifade, “Towards an emergency preparedness for self- rescue from underground coal mines,” Process Safety and Environmental Protection, vol. 149, pp. 946–957, 2021

    work page 2021

  7. [7]

    Fully convolutional networks for semantic segmentation,

    J. Long, E. Shelhamer, and T. Darrell, “Fully convolutional networks for semantic segmentation,” in 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015, pp. 3431–3440

    work page 2015

  8. [8]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” inMedical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, part III 18 . Springer, 2015, pp. 234–241

    2015

Show all 28 references
  1. [9]

    Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs,

    L.-C. Chen, G. Papandreou, I. Kokkinos, K. Murphy, and A. L. Yuille, “Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 40, no. 4, pp. 834–848, 2018

    2018

  2. [10]

    Instance segmentation in the dark,

    L. Chen, Y . Fu, K. Wei, D. Zheng, and F. Heide, “Instance segmentation in the dark,” International Journal of Computer Vision, vol. 131, no. 8, pp. 2198–2218, 2023

    2023

  3. [11]

    Advancing image understanding in poor visibility environments: A collective benchmark study,

    W. Yang et al. , “Advancing image understanding in poor visibility environments: A collective benchmark study,” IEEE Transactions on Image Processing , vol. 29, pp. 5737–5752, 2020

    2020

  4. [12]

    Kindling the darkness: A practical low-light image enhancer,

    Y . Zhang, J. Zhang, and X. Guo, “Kindling the darkness: A practical low-light image enhancer,” in Proceedings of the 27th ACM international conference on multimedia , 2019, pp. 1632– 1640

    2019

  5. [13]

    Pixelwise instance segmentation with a dynamically instantiated network,

    A. Arnab and P. H. S. Torr, “Pixelwise instance segmentation with a dynamically instantiated network,” 2017

    2017

  6. [14]

    Mask r-cnn,

    K. He, G. Gkioxari, P. Dollár, and R. Girshick, “Mask r-cnn,” in Proceedings of the IEEE international conference on computer vision, 2017, pp. 2961–2969

    2017

  7. [15]

    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,” Ad- vances in neural information processing systems , vol. 28, 2015

    2015

  8. [16]

    Hybrid task cascade for instance segmentation,

    K. Chen, J. Pang, J. Wang, Y . Xiong, X. Li, S. Sun, W. Feng, Z. Liu, J. Shi, W. Ouyang, C. C. Loy, and D. Lin, “Hybrid task cascade for instance segmentation,” 2019

    2019

  9. [17]

    Yolact: Real- time instance segmentation,

    D. Bolya, C. Zhou, F. Xiao, and Y . J. Lee, “Yolact: Real- time instance segmentation,” in Proceedings of the IEEE/CVF international conference on computer vision , 2019, pp. 9157– 9166

    2019

  10. [18]

    Insta- yolo: Real-time instance segmentation,

    E. Mohamed, A. Shaker, A. El-Sallab, and M. Hadhoud, “Insta- yolo: Real-time instance segmentation,” 2021

    2021

  11. [19]

    Microsoft coco: Common objects in context,

    T.-Y . Lin, M. Maire, S. Belongie, L. Bourdev, R. Girshick, J. Hays, P. Perona, D. Ramanan, C. L. Zitnick, and P. Dollár, “Microsoft coco: Common objects in context,” 2015

    2015

  12. [20]

    The cityscapes dataset for semantic urban scene understanding,

    M. Cordts, M. Omran, S. Ramos, T. Rehfeld, M. Enzweiler, R. Benenson, U. Franke, S. Roth, and B. Schiele, “The cityscapes dataset for semantic urban scene understanding,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 3213–3223

    2016

  13. [21]

    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,” CoRR, vol. abs/2005.12872, 2020

    2005 arXiv

  14. [22]

    Feature denoising for low-light instance segmentation using weighted non-local blocks,

    J. Lin, N. Anantrasirichai, and D. R. Bull, “Feature denoising for low-light instance segmentation using weighted non-local blocks,” ArXiv, vol. abs/2402.18307, 2024

    2024 arXiv

  15. [23]

    Segment anything,

    A. Kirillov et al. , “Segment anything,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 4015–4026

    2023

  16. [24]

    VGG image annotator (VIA),

    A. Dutta, A. Gupta, and A. Zissermann, “VGG image annotator (VIA),” http://www.robots.ox.ac.uk/ vgg/software/via/, 2016

    2016

  17. [25]

    McCann, Retinex Theory

    J. McCann, Retinex Theory . New York, NY: Springer New York, 2016, pp. 1118–1125

    2016

  18. [26]

    Orb: An efficient alternative to sift or surf,

    E. Rublee, V . Rabaud, K. Konolige, and G. Bradski, “Orb: An efficient alternative to sift or surf,” in 2011 International Conference on Computer Vision , 2011, pp. 2564–2571

    2011

  19. [27]

    Dslmf+: An open dataset for intelligent recognition of abnormal condition in underground longwall mining face,

    X. Zhang, W. Yang, B. Ma, and Y . Wang, “Dslmf+: An open dataset for intelligent recognition of abnormal condition in underground longwall mining face,” 2024

    2024

  20. [28]

    Masked-attention mask transformer for universal image segmentation,

    B. Cheng, I. Misra, A. G. Schwing, A. Kirillov, and R. Gird- har, “Masked-attention mask transformer for universal image segmentation,” arXiv, 2021

    2021

Pith tools

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