GraspSense: Physically Grounded Grasp and Grip Planning for a Dexterous Robotic Hand via Language-Guided Perception and Force Maps

Dzmitry Tsetserukou; Elizaveta Semenyakina; Ivan Snegirev; Mariya Lezina; Miguel Altamirano Cabrera; Safina Gulyamova

arxiv: 2604.05697 · v2 · submitted 2026-04-07 · 💻 cs.RO · cs.SY· eess.SY

GraspSense: Physically Grounded Grasp and Grip Planning for a Dexterous Robotic Hand via Language-Guided Perception and Force Maps

Elizaveta Semenyakina , Ivan Snegirev , Mariya Lezina , Miguel Altamirano Cabrera , Safina Gulyamova , Dzmitry Tsetserukou This is my paper

Pith reviewed 2026-05-10 19:10 UTC · model grok-4.3

classification 💻 cs.RO cs.SYeess.SY

keywords robotic graspingdexterous manipulationforce mappingphysically grounded planninggrasp selectionimpedance controlfragile object handling

0 comments

The pith

A force map computed from object geometry lets a robotic hand choose contact points that avoid weak regions and regulate grip forces safely.

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

The paper shows how to move beyond purely geometric grasp planning by adding a map of the maximum sideways force each surface point on an object can tolerate without deforming. From a spoken command the system reconstructs the object's shape, imports it into a physics simulator, and calculates this force map using the reconstructed geometry and material assumptions. Grasp candidates are first checked for geometric validity and task fit, then re-ranked to prefer contacts in stronger areas; an impedance controller then sets each finger's stiffness to match the local force limit. On paper, plastic, and glass cups the method selects structurally better contact regions and keeps forces within safe bounds. If the claim holds, dexterous hands can handle fragile everyday objects without the damage risk that geometric planners alone create.

Core claim

The central claim is that grasp selection and grip execution become a joint physically grounded problem when a force map encoding maximum admissible lateral contact force at every surface point is used both to filter and re-rank candidates and to scale finger stiffness during execution.

What carries the argument

The force map, which records the maximum lateral contact force admissible at each surface location without deformation, obtained from physics-informed geometric analysis of the reconstructed 3D model.

If this is right

Grasp candidates with comparable geometric scores are automatically re-ranked to favor mechanically stronger contact locations.
Each finger's stiffness is scaled by the admissible force value at its planned contact point, producing lower forces on weaker surface regions.
The pipeline produces consistent selection of stronger regions and grip forces that remain below damage thresholds across paper, plastic, and glass objects.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same force-map idea could be tested on objects with internal structure or varying thickness where surface geometry alone is insufficient.
Combining the map with online force sensing during the grasp might allow real-time correction if the initial material assumptions prove inaccurate.
Extending the method to two-handed or tool-using tasks would require generating and aligning force maps across multiple objects.

Load-bearing premise

The simulation accurately predicts the highest sideways force each surface point can withstand without deformation from only the object's 3D shape and assumed material properties.

What would settle it

Run the system on a paper cup whose deformation threshold is known independently, execute the selected grasp while respecting the predicted force limit, and check whether measurable deformation occurs; repeat with a grasp that ignores the force map.

Figures

Figures reproduced from arXiv: 2604.05697 by Dzmitry Tsetserukou, Elizaveta Semenyakina, Ivan Snegirev, Mariya Lezina, Miguel Altamirano Cabrera, Safina Gulyamova.

**Figure 2.** Figure 2: The proposed system transforms a natural language [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 2.** Figure 2: Overview of the proposed pipeline, comprising five stages: task understanding and semantic parsing, object perception and 3D reconstruction, [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Figure 3: Two geometrically equivalent grasps on a glass goblet (blue [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

**Figure 4.** Figure 4: Contact force vectors for the three controller conditions. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗

read the original abstract

Dexterous robotic manipulation requires more than geometrically valid grasps: it demands physically grounded contact strategies that account for the spatially non-uniform mechanical properties of the object. However, existing grasp planners typically treat the surface as structurally homogeneous, even though contact in a weak region can damage the object despite a geometrically perfect grasp. We present a pipeline for grasp selection and force regulation in a five-fingered robotic hand, based on a map of locally admissible contact loads. From an operator command, the system identifies the target object, reconstructs its 3D geometry using SAM3D, and imports the model into Isaac Sim. A physics-informed geometric analysis then computes a force map that encodes the maximum lateral contact force admissible at each surface location without deformation. Grasp candidates are filtered by geometric validity and task-goal consistency. When multiple candidates are comparable under classical metrics, they are re-ranked using a force-map-aware criterion that favors grasps with contacts in mechanically admissible regions. An impedance controller scales the stiffness of each finger according to the locally admissible force at the contact point, enabling safe and reliable grasp execution. Validation on paper, plastic, and glass cups shows that the proposed approach consistently selects structurally stronger contact regions and keeps grip forces within safe bounds. In this way, the work reframes dexterous manipulation from a purely geometric problem into a physically grounded joint planning problem of grasp selection and grip execution for future humanoid systems.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper integrates language-guided perception with simulation-derived force maps to pick safer grasps and tune impedance on a dexterous hand, but the validation is stated without numbers or calibration checks so the practical gain is unclear.

read the letter

The paper describes a grasp planning system for a dexterous hand that uses language commands to identify objects, reconstructs their shape with SAM3D, runs physics analysis in Isaac Sim to create a force map of safe contact loads, then picks grasps that avoid weak spots and adjusts finger stiffness accordingly. The validation claims success on paper, plastic, and glass cups by selecting stronger regions and staying within safe forces. What stands out is the practical combination of perception, simulation, and control to handle non-uniform object strength. Existing planners often ignore that a geometrically good grasp can still crush or tear something if it hits a weak area. Adding the force map re-ranking and impedance scaling is a direct way to address that for real objects. The approach builds on standard tools without introducing new math or algorithms, which keeps it implementable. For humanoid or service robot applications, this kind of physically grounded planning could reduce damage in everyday tasks. The main limitation is the lack of detail on the results. The abstract states the method worked but gives no numbers on success rates, force measurements, comparisons to baselines, or cases where it failed. The force map depends on material properties fed into the simulator, yet there's no indication of how those were determined or verified against the physical cups. For deformable materials like paper or thin plastic, small errors in the sim could lead to wrong rankings or unsafe forces in reality. If the full paper includes quantitative experiments, calibration steps, or real sensor data matching the sim, that would make the claims much stronger. As it stands, the central benefit rests on unshown evidence. This kind of work is useful for researchers in robotic manipulation who already use simulation pipelines and want to add safety considerations for object integrity. A reader looking for new theoretical insights won't find them here, but someone building systems might pick up the integration pattern. I would recommend sending it to peer review. The idea is solid enough and the gap it targets is real, even if the current write-up needs more experimental substance to be fully convincing.

Referee Report

2 major / 3 minor

Summary. The paper introduces GraspSense, a pipeline for dexterous grasp and grip planning that incorporates physical object properties. From a language command, it identifies the target, reconstructs 3D geometry via SAM3D, imports the model to Isaac Sim, and computes a force map encoding the maximum admissible lateral contact force at each surface point without deformation (using local geometry and assumed material properties). Grasps are filtered for geometric validity and task consistency, then re-ranked by a force-map criterion favoring mechanically admissible contact regions. An impedance controller modulates finger stiffness based on local admissible forces. The central claim is that validation on paper, plastic, and glass cups demonstrates consistent selection of structurally stronger regions while keeping grip forces within safe bounds.

Significance. If the force-map predictions prove accurate, the work would meaningfully advance robotic manipulation by reframing grasp planning as a joint geometric-plus-physical problem rather than purely geometric. The integration of language-guided perception, simulation-derived force maps, and adaptive impedance control is a coherent strength. However, the manuscript supplies no quantitative metrics, baselines, or calibration data for the simulation-based force map, so the practical significance cannot yet be assessed.

major comments (2)

[Abstract / validation experiments] Abstract and validation section: the claim that the approach 'consistently selects structurally stronger contact regions and keeps grip forces within safe bounds' on paper, plastic, and glass cups is unsupported by any reported quantitative metrics, baselines, error bars, success rates, or failure cases. This evidence is load-bearing for the headline result.
[Force map computation / §4] Force-map generation (physics-informed geometric analysis): the pipeline assumes material properties and rigid-body geometry in Isaac Sim to predict maximum admissible lateral forces, yet reports no calibration of those parameters against the physical cups nor any direct measurement of deformation or yield thresholds at candidate contact sites. For paper and thin plastic this assumption is especially brittle and directly affects both re-ranking and the safety guarantee.

minor comments (3)

[Method] Clarify the exact procedure for computing the force map from local geometry and assumed material constants; include pseudocode or equations if possible.
[Experiments] Add a table or figure showing example force maps, selected grasp candidates, and corresponding real-world outcomes for the three cup materials.
[Control] Provide implementation details for the impedance controller (e.g., how stiffness is scaled from the force-map value) and any safety margins used.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thorough review and valuable feedback on our work. We address each of the major comments in detail below and outline the revisions we plan to make to the manuscript.

read point-by-point responses

Referee: [Abstract / validation experiments] Abstract and validation section: the claim that the approach 'consistently selects structurally stronger contact regions and keeps grip forces within safe bounds' on paper, plastic, and glass cups is unsupported by any reported quantitative metrics, baselines, error bars, success rates, or failure cases. This evidence is load-bearing for the headline result.

Authors: We agree that the validation experiments in the current manuscript are primarily qualitative, demonstrating the behavior on different cup materials through example grasps and force applications. The claims in the abstract are based on these observations, but we acknowledge the lack of quantitative metrics such as success rates, force error measurements, or statistical comparisons to baselines. To address this, we will revise the validation section to include quantitative results from repeated trials, including grasp success rates, measured contact forces versus admissible limits, and comparisons against a geometric-only baseline. Error bars and failure case analysis will be added to provide a more rigorous evaluation of the headline claims. revision: yes
Referee: [Force map computation / §4] Force-map generation (physics-informed geometric analysis): the pipeline assumes material properties and rigid-body geometry in Isaac Sim to predict maximum admissible lateral forces, yet reports no calibration of those parameters against the physical cups nor any direct measurement of deformation or yield thresholds at candidate contact sites. For paper and thin plastic this assumption is especially brittle and directly affects both re-ranking and the safety guarantee.

Authors: The force map is generated using standard material properties for each category (e.g., Young's modulus for paper, plastic, and glass) within the Isaac Sim rigid-body simulation to estimate local deformation thresholds. We did not perform physical calibration or direct measurements on the specific cups used in the experiments, as the focus was on the integration of the pipeline rather than material characterization. This is a limitation, particularly for variable materials like paper. In the revised manuscript, we will explicitly state the assumed parameter values and their sources, include a discussion of the assumptions' impact, and provide a simulation-based sensitivity analysis to quantify how variations in material properties affect the force maps and grasp selection. We believe this will better contextualize the results without requiring new physical experiments at this stage. revision: partial

Circularity Check

0 steps flagged

No significant circularity; pipeline relies on external simulation and perception without self-referential reductions

full rationale

The paper describes a pipeline that reconstructs object geometry via SAM3D, imports it to Isaac Sim, computes a force map from local geometry plus assumed material properties, filters and re-ranks grasps using that map, and modulates an impedance controller accordingly. No equations, fitted parameters, or predictions are presented that reduce by construction to the inputs. The force-map computation is an external simulation step rather than a derived result internal to the paper, and validation consists of empirical checks on physical cups rather than any self-citation chain or ansatz smuggling. The derivation chain is therefore self-contained against external benchmarks and contains no load-bearing steps that equate outputs to inputs by definition.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The approach depends on unverified assumptions about 3D reconstruction accuracy and the fidelity of the physics-informed force calculation; no independent evidence for these is supplied in the abstract.

axioms (2)

domain assumption SAM3D produces a sufficiently accurate 3D geometry for subsequent physics analysis.
The pipeline imports the SAM3D output directly into Isaac Sim without stated error correction.
domain assumption The geometric analysis correctly maps surface location to maximum admissible lateral force for the given material.
This mapping is the load-bearing step that filters and re-ranks grasps.

invented entities (1)

Force map no independent evidence
purpose: Encodes maximum lateral contact force admissible at each surface point without deformation
New data structure introduced to guide grasp selection and impedance control.

pith-pipeline@v0.9.0 · 5605 in / 1420 out tokens · 70868 ms · 2026-05-10T19:10:15.880888+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

18 extracted references · 18 canonical work pages · 1 internal anchor

[1]

Trends and challenges in robot manipula- tion,

A. Billard and D. Kragic, “Trends and challenges in robot manipula- tion,”Science, vol. 364, no. 6446, p. eaat8414, 2019

work page 2019
[2]

Dexgraspnet: A large-scale robotic dexterous grasp dataset for general objects based on simulation,

R. Wang, J. Zhang, J. Chen, Y . Xu, P. Li, T. Liu, and H. Wang, “Dexgraspnet: A large-scale robotic dexterous grasp dataset for general objects based on simulation,” inProc. IEEE Int. Conf. on Robotics and Automation (ICRA), 2023, pp. 11 359–11 366

work page 2023
[3]

[Online]

Shadow Robot Company,Shadow Dexterous Hand: Technical Specification, Shadow Robot Company Ltd., 2023. [Online]. Available: https://www.shadowrobot.com/dexterous-hand-series/

work page 2023
[4]

Unidexgrasp: Universal robotic dexterous grasping via learning diverse proposal generation and goal-conditioned policy,

Y . Xu, W. Wan, J. Zhang, H. Liu, Z. Shan, H. Shen, R. Wang, H. Geng, Y . Weng, J. Chenet al., “Unidexgrasp: Universal robotic dexterous grasping via learning diverse proposal generation and goal-conditioned policy,” inProc. IEEE/CVF Conf. on Computer Vision and Pattern Recognition (CVPR), 2023, pp. 4737–4746

work page 2023
[5]

AnyDexGrasp: General dexterous grasping for different hands with human-level learning efficiency,

H.-S. Fang, H. Yan, Z. Tang, H. Fang, C. Wang, and C. Lu, “Anydex- grasp: General dexterous grasping for different hands with human-level learning efficiency,” 2025, arXiv:2502.16420

work page arXiv 2025
[6]

Grace: Balancing multiple cri- teria to achieve stable, collision-free, and functional grasps,

T. Taunyazov, K. Lin, and H. Soh, “Grace: Balancing multiple cri- teria to achieve stable, collision-free, and functional grasps,” 2024, arXiv:2309.08887

work page arXiv 2024
[7]

Adaptive motion planning for multi- fingered functional grasp via force feedback,

D. Tian, X. Lin, and Y . Sun, “Adaptive motion planning for multi- fingered functional grasp via force feedback,” inProc. IEEE-RAS Int. Conf. on Humanoid Robots (Humanoids). IEEE, Nov. 2024, p. 835–842

work page 2024
[8]

Impedance control: An approach to manipulation: Part i—theory,

N. Hogan, “Impedance control: An approach to manipulation: Part i—theory,”Journal of Dynamic Systems, Measurement, and Control, vol. 107, no. 1, pp. 1–7, 03 1985

work page 1985
[9]

A finite element based simulation framework for robotic grasp analysis,

S. Dharbaneshwer, A. Thondiyath, S. J. Subramanian, and I.-M. Chen, “A finite element based simulation framework for robotic grasp analysis,”Proc. of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science, vol. 235, no. 13, pp. 2482–2495, 2021

work page 2021
[10]

Code as policies: Language model programs for embodied control,

J. Liang, W. Huang, F. Xia, P. Xu, K. Hausman, B. Ichter, P. Florence, and A. Zeng, “Code as policies: Language model programs for embodied control,” inProc. Int. Conf. on Robotics and Automation (ICRA), 2023, pp. 9493–9500

work page 2023
[11]

Lan-grasp: Using large language models for semantic object grasping,

R. Mirjalili, M. Krawez, Y . Blei, S. Silenzi, F. Walter, and W. Burgard, “Lan-grasp: Using large language models for semantic object grasping and placement,” 2026, arXiv:2310.05239

work page arXiv 2026
[12]

Grasp as you say: Language-guided dexterous grasp generation,

Y .-L. Wei, J.-J. Jiang, C. Xing, X.-T. Tan, X.-M. Wu, H. Li, M. Cutkosky, and W.-S. Zheng, “Grasp as you say: Language-guided dexterous grasp generation,” inProc. Advances in Neural Information Processing Systems (NeurIPS), 2024

work page 2024
[13]

Task-oriented grasp prediction with visual-language inputs,

C. Tang, D. Huang, L. Meng, W. Liu, and H. Zhang, “Task-oriented grasp prediction with visual-language inputs,” inProc. IEEE/RSJ Int. Conf. on Intelligent Robots and Systems (IROS), 2023, pp. 4881–4888

work page 2023
[14]

Qwen Technical Report

J. Bai, S. Bai, Y . Chu, Z. Cui, K. Dang, X. Deng, Y . Fan, W. Ge, Y . Han, F. Huanget al., “Qwen technical report,” 2023, arXiv:2309.16609

work page internal anchor Pith review Pith/arXiv arXiv 2023
[15]

Segment anything,

A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Lo, P. Dollár, and R. Gir- shick, “Segment anything,” inProc. 2023 IEEE/CVF Int. Conf. on Computer Vision (ICCV), 2023, pp. 3992–4003

work page 2023
[16]

Yolo-world: Real-time open-vocabulary object detection,

T. Cheng, L. Song, Y . Ge, W. Liu, X. Wang, and Y . Shan, “Yolo-world: Real-time open-vocabulary object detection,”Proc. IEEE/CVF Conf. on Computer Vision and Pattern Recognition (CVPR), pp. 16 901– 16 911, 2024

work page 2024
[17]

Sam3d: Segment anything in 3d scenes.arXiv preprint arXiv:2306.03908, 2023

Y . nuo Yang, X. Wu, T. He, H. Zhao, and X. Liu, “Sam3d: Segment anything in 3d scenes,” arXiv:2306.03908

work page arXiv
[18]

Same object, different grasps: Data and semantic knowledge for task-oriented grasping,

A. Murali, W. Liu, K. Marino, S. Chernova, and A. Gupta, “Same object, different grasps: Data and semantic knowledge for task-oriented grasping,” 2020, arXiv:2011.06431

work page arXiv 2020

[1] [1]

Trends and challenges in robot manipula- tion,

A. Billard and D. Kragic, “Trends and challenges in robot manipula- tion,”Science, vol. 364, no. 6446, p. eaat8414, 2019

work page 2019

[2] [2]

Dexgraspnet: A large-scale robotic dexterous grasp dataset for general objects based on simulation,

R. Wang, J. Zhang, J. Chen, Y . Xu, P. Li, T. Liu, and H. Wang, “Dexgraspnet: A large-scale robotic dexterous grasp dataset for general objects based on simulation,” inProc. IEEE Int. Conf. on Robotics and Automation (ICRA), 2023, pp. 11 359–11 366

work page 2023

[3] [3]

[Online]

Shadow Robot Company,Shadow Dexterous Hand: Technical Specification, Shadow Robot Company Ltd., 2023. [Online]. Available: https://www.shadowrobot.com/dexterous-hand-series/

work page 2023

[4] [4]

Unidexgrasp: Universal robotic dexterous grasping via learning diverse proposal generation and goal-conditioned policy,

Y . Xu, W. Wan, J. Zhang, H. Liu, Z. Shan, H. Shen, R. Wang, H. Geng, Y . Weng, J. Chenet al., “Unidexgrasp: Universal robotic dexterous grasping via learning diverse proposal generation and goal-conditioned policy,” inProc. IEEE/CVF Conf. on Computer Vision and Pattern Recognition (CVPR), 2023, pp. 4737–4746

work page 2023

[5] [5]

AnyDexGrasp: General dexterous grasping for different hands with human-level learning efficiency,

H.-S. Fang, H. Yan, Z. Tang, H. Fang, C. Wang, and C. Lu, “Anydex- grasp: General dexterous grasping for different hands with human-level learning efficiency,” 2025, arXiv:2502.16420

work page arXiv 2025

[6] [6]

Grace: Balancing multiple cri- teria to achieve stable, collision-free, and functional grasps,

T. Taunyazov, K. Lin, and H. Soh, “Grace: Balancing multiple cri- teria to achieve stable, collision-free, and functional grasps,” 2024, arXiv:2309.08887

work page arXiv 2024

[7] [7]

Adaptive motion planning for multi- fingered functional grasp via force feedback,

D. Tian, X. Lin, and Y . Sun, “Adaptive motion planning for multi- fingered functional grasp via force feedback,” inProc. IEEE-RAS Int. Conf. on Humanoid Robots (Humanoids). IEEE, Nov. 2024, p. 835–842

work page 2024

[8] [8]

Impedance control: An approach to manipulation: Part i—theory,

N. Hogan, “Impedance control: An approach to manipulation: Part i—theory,”Journal of Dynamic Systems, Measurement, and Control, vol. 107, no. 1, pp. 1–7, 03 1985

work page 1985

[9] [9]

A finite element based simulation framework for robotic grasp analysis,

S. Dharbaneshwer, A. Thondiyath, S. J. Subramanian, and I.-M. Chen, “A finite element based simulation framework for robotic grasp analysis,”Proc. of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science, vol. 235, no. 13, pp. 2482–2495, 2021

work page 2021

[10] [10]

Code as policies: Language model programs for embodied control,

J. Liang, W. Huang, F. Xia, P. Xu, K. Hausman, B. Ichter, P. Florence, and A. Zeng, “Code as policies: Language model programs for embodied control,” inProc. Int. Conf. on Robotics and Automation (ICRA), 2023, pp. 9493–9500

work page 2023

[11] [11]

Lan-grasp: Using large language models for semantic object grasping,

R. Mirjalili, M. Krawez, Y . Blei, S. Silenzi, F. Walter, and W. Burgard, “Lan-grasp: Using large language models for semantic object grasping and placement,” 2026, arXiv:2310.05239

work page arXiv 2026

[12] [12]

Grasp as you say: Language-guided dexterous grasp generation,

Y .-L. Wei, J.-J. Jiang, C. Xing, X.-T. Tan, X.-M. Wu, H. Li, M. Cutkosky, and W.-S. Zheng, “Grasp as you say: Language-guided dexterous grasp generation,” inProc. Advances in Neural Information Processing Systems (NeurIPS), 2024

work page 2024

[13] [13]

Task-oriented grasp prediction with visual-language inputs,

C. Tang, D. Huang, L. Meng, W. Liu, and H. Zhang, “Task-oriented grasp prediction with visual-language inputs,” inProc. IEEE/RSJ Int. Conf. on Intelligent Robots and Systems (IROS), 2023, pp. 4881–4888

work page 2023

[14] [14]

Qwen Technical Report

J. Bai, S. Bai, Y . Chu, Z. Cui, K. Dang, X. Deng, Y . Fan, W. Ge, Y . Han, F. Huanget al., “Qwen technical report,” 2023, arXiv:2309.16609

work page internal anchor Pith review Pith/arXiv arXiv 2023

[15] [15]

Segment anything,

A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Lo, P. Dollár, and R. Gir- shick, “Segment anything,” inProc. 2023 IEEE/CVF Int. Conf. on Computer Vision (ICCV), 2023, pp. 3992–4003

work page 2023

[16] [16]

Yolo-world: Real-time open-vocabulary object detection,

T. Cheng, L. Song, Y . Ge, W. Liu, X. Wang, and Y . Shan, “Yolo-world: Real-time open-vocabulary object detection,”Proc. IEEE/CVF Conf. on Computer Vision and Pattern Recognition (CVPR), pp. 16 901– 16 911, 2024

work page 2024

[17] [17]

Sam3d: Segment anything in 3d scenes.arXiv preprint arXiv:2306.03908, 2023

Y . nuo Yang, X. Wu, T. He, H. Zhao, and X. Liu, “Sam3d: Segment anything in 3d scenes,” arXiv:2306.03908

work page arXiv

[18] [18]

Same object, different grasps: Data and semantic knowledge for task-oriented grasping,

A. Murali, W. Liu, K. Marino, S. Chernova, and A. Gupta, “Same object, different grasps: Data and semantic knowledge for task-oriented grasping,” 2020, arXiv:2011.06431

work page arXiv 2020