Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Measure Anything: Real-time, Multi-stage Vision-based Dimensional Measurement using Segment Anything

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

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

Pith's one-line read This paper claims that a multi-stage pipeline built on SAM 2 can convert one RGB-D frame into continuous diameter, length, and volume measurements for circular-cross-section objects, and that these measurements can support automated crop…

desk verdict A coherent SAM2-based measurement pipeline whose core diameter accuracy claim is unverified; the missing ground-truth experiment, not the geometry, is the load-bearing gap. read the letter →

arxiv 2412.03472 v1 pith:GD4HBITE submitted 2024-12-04 cs.CV

Yongkyu Lee , Shivam Kumar Panda , Wei Wang , Mohammad Khalid Jawed This is my paper · ORCID
classification cs.CV
keywords dimensionalmeasurementSegmentAnythingModelRGB-Dskeletonconstructionstemdiametercropphenotypingroboticgraspingcircularcross-section
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

Measure Anything claims that one RGB-D frame plus point prompts is enough to extract a continuous dimensional profile — diameter at many stations, total length, and volume — for objects with circular cross-sections, including curved rods and near-symmetric general shapes. It builds on the Segment Anything Model (SAM 2) for segmentation, then refines the mask, constructs an ordered skeleton from bottom to top, draws line segments perpendicular to the local skeleton slope, and back-projects the segment endpoints to 3D using depth. The paper validates the pipeline on Canola stems from North Dakota, where measuring stem diameter matters for phenotyping, and shows the measurements stay within roughly ±10 percent of ground truth for a bottle viewed from varying camera positions. It also demonstrates that the same diameter profile can feed a simple grasp stability model, selecting grasp points near concave surfaces and the center of gravity.

What carries the argument

The central mechanism is an ordered skeleton with perpendicular line segments and median-depth 3D back-projection. For rod-like objects the skeleton comes from a medial axis transform, pruned by removing short branches, then reordered bottom-to-top so adjacent pixels are connected; for general shapes a principal-component axis is selected by a reflection dissimilarity test. At each sampled skeleton pixel the local slope is computed by central differences, and the algorithm walks outward in the perpendicular direction until it leaves the mask, recording the endpoint and collecting depth values along the ray; the median depth of each side handles missing depth near contours. The endpoints' 3D coordinates are then used directly as diameters, and the skeleton ordering makes length and volume simple sums.

What would settle it

Take a straight cylinder of known diameter, film it with a stereo depth camera at known tilts (0°, 15°, 30°, 45°) toward or away from the camera, run the full pipeline, and compare the estimated mid-stem diameter with the ground truth; if the estimate shifts systematically with tilt angle instead of staying constant, the perpendicular-segment assumption is the source of the error. The predicted chord length for a known cylinder radius and tilt can be computed analytically and checked against the pipeline output.

Watch

Extended reading notes

Core claim

The central claim is that dimensional measurement of rod-like and near-symmetric objects reduces to a sequence of well-defined geometry operations on a binary mask plus depth map: remove mask noise, compute a single ordered skeleton, sample it, find the mask boundary along the direction perpendicular to the local skeleton slope, take the median depth on each side of the boundary segment, and project the 2D endpoints into 3D with camera intrinsics. The kth diameter is D_k = ||X_{k1} - X_{k2}||, the Euclidean distance between the projected endpoints; length is the sum of distances between consecutive 3D midpoints; volume is the sum of truncated-cone volumes between consecutive segment pairs. This converts the segmentation power of SAM 2 into concrete physical quantities without training a measurement network, and the paper reports that combining the pipeline with a keypoint detector automates Canola stem diameter estimation while an interactive version lets a user click objects in cluttered scenes.

Load-bearing premise

The whole measurement hinges on treating the image-perpendicular line through the skeleton as a true circular cross-section of the object; that is exact only when the object's axis is parallel to the image plane, and tilted rods make the line an off-center chord whose measured width can be wrong.

Editorial extensions

If this is right

  • Continuous diameter profiles enable trait extraction such as taper or maximum and minimum stem width from a single frame, not just an average diameter.
  • With a keypoint detector supplying prompts, the pipeline can run unattended over many images, supporting high-throughput field phenotyping.
  • The same measured profile can be piped into grasp scoring: the stability model ranks segments near concave surfaces and close to the projected center of gravity.
  • Length and volume remain within roughly ±10 percent across camera positions in the authors' bottle experiment, suggesting the measurements are usable when depth quality is reasonable.
  • Because the framework inherits SAM 2's promptable masks, users can correct mistakes by adding positive or negative points instead of retraining.

Reading between the lines

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

  • An untested extension is to estimate the local 3D skeleton orientation from depth and correct for foreshortening, which should remove tilt bias from the diameter at little computational cost.
  • Volume is integrated from the diameter profile, so any systematic diameter error from tilt would compound in the volume estimate; the paper's ±10 percent result may degrade for strongly tilted objects.
  • The same mask-plus-skeleton machinery could measure other widths, such as crack openings or tube bores, by changing which boundary points are paired.
  • Adding SAM 2's video mode could turn the pipeline into a non-contact deformation sensor for soft rods, reporting diameter and curvature over time.
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 / 5 minor

Summary. The paper presents 'Measure Anything,' a modular pipeline that combines SAM 2 segmentation with mask processing, skeleton construction, perpendicular line-segment extraction, and 2D-3D transformation to estimate diameter, length, and volume of objects with circular cross-sections. Two prompt modes are described: interactive point prompts and automated keypoint prompts from a YOLOv8 detector. The framework is demonstrated on Canola stem phenotyping and on robotic grasp-point selection using the Clubs dataset. Quantitative validation consists of a wine-bottle experiment reporting length and volume within ±10% of manual ground truth over roughly 90 frames, an 83.5% mAP score for the keypoint detector, and a qualitative grasp-stability visualization.

Significance. If the measurement accuracy claims are substantiated, the framework would be a useful open-source integration of foundation-model segmentation with classical geometric measurement, potentially benefiting precision agriculture and robotic manipulation. The authors provide a public code and dataset repository and a clear modular design. However, the central claim of 'accurate' dimensional measurement, particularly diameter estimation, is not yet supported by the reported experiments: no ground-truth diameter comparisons, error metrics, or sample-size statistics are provided for the target application. The work is therefore a promising systems contribution whose main advertised quantity remains unvalidated.

major comments (4)
  1. [Section IV, Fig. 4] The paper's abstract and contribution list claim validation of Canola stem diameter estimation, but Section IV reports no quantitative diameter comparisons. The only numeric validation is the wine-bottle experiment, which reports length and volume variation within ±10% and does not report diameter error. The authors should add a table or plot comparing estimated diameters with caliper or other ground truth for Canola stems, including per-frame statistics, sample size, and error metrics such as MAE, RMSE, or bias. Without this, the central claim that the pipeline 'accurately measures' diameter is unsupported.
  2. [Section III-D, Algorithm 1] The depth filter in Algorithm 1 (line 14) accepts only depths satisfying |d - d0| < 0.1d0 relative to the depth at the skeleton pixel, and it stops the outward search at the first invalid depth. If the object is tilted or the depth map is noisy, this can truncate the line segment before reaching the true contour, directly biasing D_k. The manuscript provides no sensitivity analysis for the 10% threshold or for missing-depth behavior. The authors should report how often endpoints were truncated, quantify the resulting diameter bias, or justify the threshold empirically.
  3. [Section III-G, Eq. (1)] The grasp stability model in Eq. (1) uses free weights w1-w3 and defines 'cond' only as 'the condition if the pair of coordinates lie at a minima of a concave surface.' No calibration, ablation, or quantitative grasp-success evaluation is provided; the claim that the top-scoring segments are 'optimal' is supported only by a qualitative figure and intuitive agreement. The authors should at least report a comparison with a baseline grasp heuristic or physical grasp trials, or explicitly reframe this section as a qualitative demonstration.
  4. [Section III-E, volume formula] The volume formula sums frustum volumes between consecutive line segments, but the manuscript does not state how missing or rejected line segments (due to depth filtering) affect the indexing of D_k and the arc-length l_k. If some segments are skipped, the formula silently assumes equal spacing along the skeleton, which may bias volume estimates. Please clarify how n' and the l_k values are computed when line segments are rejected, and whether this occurs in the reported wine-bottle experiment.
minor comments (5)
  1. [Abstract and Section IV] The abstract states that the framework is validated by estimating Canola stem diameters, but the results section contains no quantitative diameter analysis for Canola stems. Please align the abstract with the actual reported experiments or add the missing validation.
  2. [Fig. 4 caption] The caption ends with a stray slash: 'camera positions/'. This should be corrected.
  3. [Section III-G, Eq. (1)] The term 'cond' is described as 'the condition if the pair of coordinates lie at a minima of a concave surface.' This is ambiguous: is cond a binary indicator, a continuous curvature measure, or a condition number? Please define it precisely and state the units or normalization.
  4. [Section III-B] The mask-processing parameters (minimum size threshold, morphological kernel size) are not specified or tuned in the experiments. Reporting the chosen values and their effect on downstream measurements would improve reproducibility.
  5. [Section IV] The keypoint detector's mAP is reported as 'mAP-95' without clarifying the IoU threshold or averaging convention. Please use standard notation such as mAP@0.5:0.95 or specify the metric used.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the measurement equations are geometric definitions applied to new data, and the validation is against external manual ground truth.

full rationale

The derivation chain is self-contained. Diameter is defined as the 3D Euclidean distance between endpoints of a mask-constrained perpendicular segment (D_k = ||X_{k1} - X_{k2}||), length is the sum of consecutive 3D midpoint distances, and volume is the frustum sum over those segments. None of these quantities is fitted to the reported outputs, and the formulas are not defined in terms of the target measurements. The only external validation reported, the wine-bottle length and volume comparison with manually measured ground truth (29.845 cm and 943 ml), is independent of the pipeline's equations. The keypoint detection model is validated by mAP-95 (83.5%) on annotated images, not by the dimensional outputs, so it is not a fitted parameter renamed as a prediction. The stability weights w1-w3 in Eq. (1) are used only for the grasping demonstration, not for the dimensional measurement claim. Reference [34] is a same-group algorithmic subroutine for endpoint and intersection identification, but it is not load-bearing to the central measurement claim and does not constrain the result. The absence of reported diameter ground-truth accuracy is an evidence or validation gap, not a circularity.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The pipeline's accuracy rests on a set of geometric and sensing assumptions, plus several unspecified tuning parameters. The most consequential is that the image-perpendicular line segment is a 3D diameter of a circular cross-section; the other assumptions concern calibration, mask cleanup, and skeleton validity. No new physical entities are introduced.

free parameters (6)
  • Morphological kernel size
    Used for opening and closing in Section III-B; size is not specified and affects boundary smoothness.
  • Sampling stride p1
    Determines spacing of line segments along the skeleton in Section III-D; not specified.
  • Slope offset p2
    Number of pixels used for central-difference slope in Section III-D; not specified.
  • Depth consistency threshold = 0.1 * d0
    Hard-coded relative threshold in Algorithm 1 line 14 that determines which depth samples are kept; affects endpoint depths.
  • Stability weights w1, w2, w3 = not reported
    Weights in Eq. (1) for the grasping demo; no values, grid, or sensitivity analysis are given.
  • Short-branch pruning threshold = maximum diameter from MAT
    Threshold d in Section III-C classifies skeleton branches as artifacts; value is derived from the mask, not externally calibrated.
assumptions (5)
  • domain assumption Target objects have circular cross-sections and the line perpendicular to the 2D skeleton maps to a diametral chord in 3D.
    Statement in Section III-E and conclusion ('current pipeline assumes objects with circular cross-sections'); the diameter formula D_k = ||X_{k1} - X_{k2}|| is only exact when the image-perpendicular plane passes through the axis, which fails for strongly tilted rods unless corrected.
  • domain assumption Depth map is spatially aligned with RGB and calibrated with intrinsics and distortion coefficients.
    Section III-E invokes 'the intrinsics and distortion coefficients for the RGB and depth image' without quantifying calibration error.
  • domain assumption The largest connected component after mask processing is the target object.
    Section III-B discards all but the largest component, so any occluded or split target is mis-measured.
  • domain assumption The PCA-based skeleton axis for general geometries is the symmetry axis.
    Section III-C selects the principal axis with lower reflection dissimilarity; for non-symmetric objects this still imposes a straight skeleton with 'constant skeleton slope'.
  • domain assumption Median depth over the sampled ray approximates the true 3D location of contour endpoints.
    Algorithm 1 collects depths while |d-d0| < 0.1d0 and takes the median, assuming the object surface is continuous and near-fronto-parallel.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Measure Anything: Real-time, Multi-stage Vision-based Dimensional Measurement using Segment Anything." pith.science (2026). https://pith.science/paper/GD4HBITE

@misc{pith2026241203472,
  author       = {Pith},
  title        = {Pith review of: Measure Anything: Real-time, Multi-stage Vision-based Dimensional Measurement using Segment Anything},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GD4HBITE}},
  note         = {Machine review of arXiv:2412.03472}
}
read the original abstract

We present Measure Anything, a comprehensive vision-based framework for dimensional measurement of objects with circular cross-sections, leveraging the Segment Anything Model (SAM). Our approach estimates key geometric features -- including diameter, length, and volume -- for rod-like geometries with varying curvature and general objects with constant skeleton slope. The framework integrates segmentation, mask processing, skeleton construction, and 2D-3D transformation, packaged in a user-friendly interface. We validate our framework by estimating the diameters of Canola stems -- collected from agricultural fields in North Dakota -- which are thin and non-uniform, posing challenges for existing methods. Measuring its diameters is critical, as it is a phenotypic traits that correlates with the health and yield of Canola crops. This application also exemplifies the potential of Measure Anything, where integrating intelligent models -- such as keypoint detection -- extends its scalability to fully automate the measurement process for high-throughput applications. Furthermore, we showcase its versatility in robotic grasping, leveraging extracted geometric features to identify optimal grasp points.

Figures

Figures reproduced from arXiv: 2412.03472 by the authors.

Figure 1
Figure 1. Overview of the Measure Anything framework [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Skeleton Construction and Line Segment Depth Identification Modules. (a) Workflow for skeleton construction modules tailored to rod-like and [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Demonstration of Measure Anything on Canola stems using the interactive, automated method. (a) Interactive method requires any number of [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Variation in length and volume measurements of the object observed [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Diameter analysis using Measure Anything for identifying optimal [PITH_FULL_IMAGE:figures/full_fig_p007_5.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. Digital measurement of droplet flame diameter in microgravity combustion images using Segment Anything Model 2 with automatic prompt selection

    cs.CV 2026-07 conditional novelty 5.0 of 10

    An automated SAM2-based pipeline measures droplet flame diameter in microgravity combustion videos with about 3.1% average error versus manual reference and a 229x speedup.

Reference graph

Works this paper leans on

38 extracted references · 24 canonical work pages · cited by 1 Pith paper

  1. [1]

    Segment anything

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer Whitehead, Alexan- der C Berg, Wan-Yen Lo, et al. Segment anything. arXiv preprint arXiv:2304.02643, 2023

    arXiv 2023

  2. [2]

    Sam 2: Segment anything in images and videos

    Nikhila Ravi, Valentin Gabeur, Yuan-Ting Hu, Ronghang Hu, Chai- tanya Ryali, Tengyu Ma, Haitham Khedr, Roman R ¨adle, Chloe Rol- land, Laura Gustafson, et al. Sam 2: Segment anything in images and videos. arXiv preprint arXiv:2408.00714 , 2024

    arXiv 2024

  3. [3]

    Stem diameter variations as a versatile research tool in ecophysiology

    Tom De Swaef, Veerle De Schepper, Maurits W Vandegehuchte, and Kathy Steppe. Stem diameter variations as a versatile research tool in ecophysiology. Tree Physiology, 35(10):1047–1061, 2015

    work page 2015

  4. [4]

    P-agslam: In-row and under-canopy slam for agricultural monitoring in cornfields

    Kitae Kim, Aarya Deb, and David J Cappelleri. P-agslam: In-row and under-canopy slam for agricultural monitoring in cornfields. IEEE Robotics and Automation Letters , 2024

    work page 2024

  5. [5]

    Rgb-d salient object detection: A survey

    Tao Zhou, Deng-Ping Fan, Ming-Ming Cheng, Jianbing Shen, and Ling Shao. Rgb-d salient object detection: A survey. Computational Visual Media, 7:37–69, 2021

    work page 2021

  6. [6]

    Development of structured light 3d-scanner with high spatial resolution and its applications for additive manufacturing qual- ity assurance

    Rongxuan Wang, Andrew C Law, David Garcia, Shuo Yang, and Zhenyu Kong. Development of structured light 3d-scanner with high spatial resolution and its applications for additive manufacturing qual- ity assurance. The International Journal of Advanced Manufacturing Technology, 117:845–862, 2021

    work page 2021

  7. [7]

    Measurement meth- ods of 3d shape of large-scale complex surfaces based on computer vision: A review

    Hang Shang, Changying Liu, and Ruijian Wang. Measurement meth- ods of 3d shape of large-scale complex surfaces based on computer vision: A review. Measurement, 197:111302, 2022

    work page 2022

  8. [8]

    Fruit detection, segmentation and 3d visualisation of environments in apple orchards

    Hanwen Kang and Chao Chen. Fruit detection, segmentation and 3d visualisation of environments in apple orchards. Computers and Electronics in Agriculture, 171:105302, 2020

    work page 2020

Show all 38 references
  1. [9]

    On-tree apple fruit size estimation using stereo vision with deep learning-based occlusion handling

    Omeed Mirbod, Daeun Choi, Paul H Heinemann, Richard P Marini, and Long He. On-tree apple fruit size estimation using stereo vision with deep learning-based occlusion handling. Biosystems Engineering, 226:27–42, 2023

    2023

  2. [10]

    Length phenotyping with interest point detection

    Adar Vit, Guy Shani, and Aharon Bar-Hillel. Length phenotyping with interest point detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops , pages 0–0, 2019

    2019

  3. [11]

    Measuring stem diameter of sorghum plants in the field using a high-throughput stereo vision system

    Lirong Xiang, Lie Tang, Jingyao Gai, and Le Wang. Measuring stem diameter of sorghum plants in the field using a high-throughput stereo vision system. Transactions of the ASABE , 64(6):1999–2010, 2021

    1999

  4. [12]

    Mask r-cnn

    Kaiming He, Georgia Gkioxari, Piotr Doll ´ar, and Ross Girshick. Mask r-cnn. In Proceedings of the IEEE international conference on computer vision , pages 2961–2969, 2017

    2017

  5. [13]

    Stereo processing by semiglobal matching and mutual information

    Heiko Hirschmuller. Stereo processing by semiglobal matching and mutual information. IEEE Transactions on pattern analysis and machine intelligence, 30(2):328–341, 2007

    2007

  6. [14]

    Computer-vision-based sensing technologies for livestock body dimension measurement: A survey

    Weihong Ma, Yi Sun, Xiangyu Qi, Xianglong Xue, Kaixuan Chang, Zhankang Xu, Mingyu Li, Rong Wang, Rui Meng, and Qifeng Li. Computer-vision-based sensing technologies for livestock body dimension measurement: A survey. Sensors, 24(5):1504, 2024

    2024

  7. [15]

    Image analysis and com- puter vision applications in animal sciences: an overview

    Arthur Francisco Ara ´ujo Fernandes, Jo ˜ao Ricardo Rebouc ¸as D´orea, and Guilherme Jord ˜ao de Magalh ˜aes Rosa. Image analysis and com- puter vision applications in animal sciences: an overview. Frontiers in Veterinary Science, 7:551269, 2020

    2020

  8. [16]

    Robotic grasping and contact: A review

    Antonio Bicchi and Vijay Kumar. Robotic grasping and contact: A review. In Proceedings 2000 ICRA. Millennium conference. IEEE international conference on robotics and automation. Symposia pro- ceedings (Cat. No. 00CH37065) , volume 1, pages 348–353. IEEE, 2000

    2000

  9. [17]

    Comprehensive review on reaching and grasping of objects in robotics

    Qaid Mohammed Marwan, Shing Chyi Chua, and Lee Chung Kwek. Comprehensive review on reaching and grasping of objects in robotics. Robotica, 39(10):1849–1882, 2021

    2021

  10. [18]

    State-of-the-art robotic grippers, grasping and control strategies, as well as their applications in agricultural robots: A review

    Baohua Zhang, Yuanxin Xie, Jun Zhou, Kai Wang, and Zhen Zhang. State-of-the-art robotic grippers, grasping and control strategies, as well as their applications in agricultural robots: A review. Computers and Electronics in Agriculture , 177:105694, 2020

    2020

  11. [19]

    Vision- based robotic grasping from object localization, object pose estimation to grasp estimation for parallel grippers: a review

    Guoguang Du, Kai Wang, Shiguo Lian, and Kaiyong Zhao. Vision- based robotic grasping from object localization, object pose estimation to grasp estimation for parallel grippers: a review. Artificial Intelli- gence Review, 54(3):1677–1734, 2021

    2021

  12. [20]

    A vision-based robot grasping system

    Hu Cheng, Yingying Wang, and Max Q-H Meng. A vision-based robot grasping system. IEEE Sensors Journal , 22(10):9610–9620, 2022

    2022

  13. [21]

    Graspnet- 1billion: A large-scale benchmark for general object grasping

    Hao-Shu Fang, Chenxi Wang, Minghao Gou, and Cewu Lu. Graspnet- 1billion: A large-scale benchmark for general object grasping. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11444–11453, 2020

    2020

  14. [22]

    Graspness discovery in clutters for fast and accurate grasp detection

    Chenxi Wang, Hao-Shu Fang, Minghao Gou, Hongjie Fang, Jin Gao, and Cewu Lu. Graspness discovery in clutters for fast and accurate grasp detection. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 15964–15973, 2021

    2021

  15. [23]

    Anygrasp: Robust and efficient grasp perception in spatial and temporal domains

    Hao-Shu Fang, Chenxi Wang, Hongjie Fang, Minghao Gou, Jirong Liu, Hengxu Yan, Wenhai Liu, Yichen Xie, and Cewu Lu. Anygrasp: Robust and efficient grasp perception in spatial and temporal domains. IEEE Transactions on Robotics , 2023

    2023

  16. [24]

    Image segmentation using deep learning: A survey

    Shervin Minaee, Yuri Boykov, Fatih Porikli, Antonio Plaza, Nasser Kehtarnavaz, and Demetri Terzopoulos. Image segmentation using deep learning: A survey. IEEE transactions on pattern analysis and machine intelligence, 44(7):3523–3542, 2021

    2021

  17. [25]

    Fully convo- lutional networks for semantic segmentation

    Jonathan Long, Evan Shelhamer, and Trevor Darrell. Fully convo- lutional networks for semantic segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 3431–3440, 2015

    2015

  18. [26]

    Learning deconvolution network for semantic segmentation

    Hyeonwoo Noh, Seunghoon Hong, and Bohyung Han. Learning deconvolution network for semantic segmentation. In Proceedings of the IEEE international conference on computer vision , pages 1520– 1528, 2015

    2015

  19. [27]

    Segnet: A deep convolutional encoder-decoder architecture for image segmenta- tion

    Vijay Badrinarayanan, Alex Kendall, and Roberto Cipolla. Segnet: A deep convolutional encoder-decoder architecture for image segmenta- tion. IEEE transactions on pattern analysis and machine intelligence , 39(12):2481–2495, 2017

    2017

  20. [28]

    U-net: Con- volutional networks for biomedical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Con- volutional networks for biomedical image segmentation. In Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, par...

    2015

  21. [29]

    3d u-net: learning dense volumetric segmen- tation from sparse annotation

    ¨Ozg¨un C ¸ ic ¸ek, Ahmed Abdulkadir, Soeren S Lienkamp, Thomas Brox, and Olaf Ronneberger. 3d u-net: learning dense volumetric segmen- tation from sparse annotation. In Medical Image Computing and Computer-Assisted Intervention–MICCAI 2016: 19th International Conference, Athe...

    2016

  22. [30]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2020

    2010 arXiv

  23. [31]

    End-to-end object detec- tion with transformers

    Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, and Sergey Zagoruyko. End-to-end object detec- tion with transformers. In European conference on computer vision , pages 213–229. Springer, 2020

    2020

  24. [32]

    Mask dino: Towards a unified transformer-based framework for object detection and segmentation

    Feng Li, Hao Zhang, Huaizhe Xu, Shilong Liu, Lei Zhang, Li- onel M Ni, and Heung-Yeung Shum. Mask dino: Towards a unified transformer-based framework for object detection and segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pa...

    2023

  25. [33]

    A transformation for extracting new descriptions of shape

    Harry Blum. A transformation for extracting new descriptions of shape. Models for the perception of speech and visual form , pages 362–380, 1967

    1967

  26. [34]

    mbest: Realtime deformable linear object detection through minimal bending energy skeleton pixel traversals

    Andrew Choi, Dezhong Tong, Brian Park, Demetri Terzopoulos, Jungseock Joo, and Mohammad Khalid Jawed. mbest: Realtime deformable linear object detection through minimal bending energy skeleton pixel traversals. arXiv preprint arXiv:2302.09444 , 2023

    2023 arXiv

  27. [35]

    Grounding dino: Marrying dino with grounded pre-training for open- set object detection

    Shilong Liu, Zhaoyang Zeng, Tianhe Ren, Feng Li, Hao Zhang, Jie Yang, Qing Jiang, Chunyuan Li, Jianwei Yang, Hang Su, et al. Grounding dino: Marrying dino with grounded pre-training for open- set object detection. arXiv preprint arXiv:2303.05499 , 2023

    2023 arXiv

  28. [36]

    Clubs: An rgb-d dataset with cluttered box scenes containing household objects

    Tonci Novkovic, Fadri Furrer, Marko Panjek, Margarita Grinvald, Roland Siegwart, and Juan Nieto. Clubs: An rgb-d dataset with cluttered box scenes containing household objects. The International Journal of Robotics Research , 38(14):1538–1548, 2019

    2019

  29. [37]

    Mvgrasp: Real- time multi-view 3d object grasping in highly cluttered environments

    Hamidreza Kasaei and Mohammadreza Kasaei. Mvgrasp: Real- time multi-view 3d object grasping in highly cluttered environments. Robotics and Autonomous Systems , 160:104313, 2023

    2023

  30. [38]

    On the closure properties of robotic grasping

    Antonio Bicchi. On the closure properties of robotic grasping. The International Journal of Robotics Research , 14(4):319–334, 1995

    1995

Pith tools

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