Pith. sign in

REVIEW 4 major objections 5 minor 51 references

Robot Learning to Communicate through Projected Visual Abstractions

T0 review · 4 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read A soft-skinned robot hand shapes its own shadow into letters, puppets, and animal forms by learning a differentiable self-model of how its 21 joints project onto a 2D silhouette.

desk verdict Genuinely new capability with a sound simulation-side pipeline; sim-to-real shadow fidelity is the load-bearing unquantified assumption. read the letter →

arxiv 2607.22434 v1 pith:NRVYOTRQ submitted 2026-07-24 cs.RO cs.AI

Danyang Yan , Boyuan Wang , Jiaxun Liu , Boyuan Chen This is my paper · ORCID
classification cs.ROcs.AI
keywords projectedvisualabstractionsshadowself-modeldexterousrobothandsoftfoamskindifferentiableoptimizationcollision-awarerefinementexpressiveregionstemporalcoherence
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper attempts to establish that robots can communicate through projected visual abstractions—not just through their bodies, but through transformed representations of their bodies such as shadows. To show this, the authors build a 21-degree-of-freedom robot hand whose soft foam skin seals the gaps between finger links, producing continuous silhouettes, and train a shadow self-model that maps joint angles to the resulting binary shadow image. Given a target silhouette, the robot searches over poses by backpropagating image and perceptual losses through this forward model, then refines the pose with a collision-aware hill-climbing step so it is executable on hardware. For moving targets, the paper adds expressive-region weighting, temporal smoothing, and keyframe clustering so that dynamic performances preserve perceptually important details without optimizing every frame. The reported result is that the system reproduces ASL hand signs, hand-shadow puppetry, and animal-motion silhouettes in simulation and on the physical hand, suggesting that projected shadows can be a general communication medium for robots.

What carries the argument

The shadow self-model is the central object: an analytical forward-kinematics layer, which turns a 21-dimensional joint vector into homogeneous transforms of the finger links, feeding a neural decoder that outputs a 256×256 binary shadow image. Because every step is differentiable, the model converts the problem of pose search into gradient descent on the input joint angles. The hardware counterpart is the soft foamed-TPU skin around the rigid skeleton, which blocks light between fingers and makes continuous silhouettes physically possible; the paper separates the collision mesh (skeleton) from the visual mesh (skin envelope) in simulation, which is what lets the optimizer reason about light

What would settle it

Set up the same spotlight-backdrop-camera geometry as the paper, pick a handful of poses that require finger-on-finger contact (e.g., the duck-beak or deer-horns targets), render the predicted shadow in the simulator, then photograph the physical robot's shadow in the same pose; if the mean intersection-over-union between the binarized predicted and real shadows falls below roughly 0.8, or if enclosed regions like eyes systematically disappear, the sim-to-real shadow fidelity premise is falsified.

Watch

Extended reading notes

Core claim

The central claim is that the many-to-one relationship between a robot's 3D configuration and its projected 2D shadow is not a barrier to control; it is a handle. Instead of learning the ill-posed inverse map from shadow to pose, the authors learn a differentiable forward map—the shadow self-model—from joint angles to shadow appearance, and then optimize over poses by gradient descent through that map. A physics-based local refinement step restores physical feasibility while preserving the matched silhouette. The paper argues that this forward-model-plus-refinement strategy outperforms inverse prediction, nearest-neighbor retrieval, and pure stochastic search, and that the same pipeline, ext

Load-bearing premise

The paper assumes that the simulated visual mesh of the soft foam skin predicts, with high fidelity, how the real compliant skin deforms, compresses between fingers, and blocks light; if real shadows differ from simulated ones in these contact regions, the optimized poses will not produce the target silhouettes on the physical robot, and the claimed capability rests on unverified visual realism.

Editorial extensions

If this is right

  • A robot can acquire a shadow self-model entirely through its own simulated motor exploration, with no human demonstrations or video training data, and then use it to imitate arbitrary silhouettes at deployment.
  • Gradient-based search through a learned forward model, followed by a short collision-aware local search, yields executable poses in far fewer steps than pure hill climbing, while preserving fine features such as eyes and enclosed regions.
  • Emphasizing moving and enclosed regions during optimization—rather than matching the whole silhouette uniformly—substantially improves the perceived expressiveness of dynamic shadow performances.
  • Optimizing video sequences with inherited initialization from the previous frame raises the temporal transition success rate from roughly 34% to 86% in the paper's evaluation, meaning the same finger consistently represents the same semantic part across frames.
  • Reducing target videos to a sparse set of keyframes cuts the number of optimization targets by 63–92% while retaining the characteristic motion, so dynamic shadow imitation is computationally tractable.

Reading between the lines

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

  • If the sim-to-real shadow fidelity holds beyond the paper's qualitative demonstrations, the same forward-self-model recipe could be transferred to other robot morphologies—legged robots, manipulator arms, or soft grippers—since the model is learned from the robot's own data rather than from annotated human examples.
  • Because the self-model is differentiable, a natural next step would be closed-loop shadow control: compare the observed shadow to the target during execution and adjust joint angles online, which the paper notes is currently offline-only; that would enable interactive or adaptive performances under changing lighting.
  • The expressive-region mask is currently derived from motion differences and enclosed holes; a stronger editorial hypothesis is that a learned visual-saliency or perceptual-embedding model could define these regions automatically, possibly matching human judgments of what makes a shadow 'expressive' more closely.
  • The framework treats the shadow as a binary silhouette; an extension to grayscale shadows with controllable light-source position would add shading information, making the inverse problem more constrained and potentially enabling richer communication channels.
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 a framework for enabling a 21-DoF dexterous robotic hand with compliant soft skin to intentionally manipulate its projected shadow, treating the shadow as a visual communication medium. The system first trains a differentiable 'shadow self-model' in simulation, mapping joint configurations to binary shadow images via analytical forward kinematics followed by a neural decoder. At deployment, a target shadow image or video is reproduced by gradient-based search over the frozen self-model, followed by collision-aware hill-climbing refinement in a physics simulator. For video targets, the method adds expressive-region objectives, temporal smoothness regularization, inherited initialization, and PCA-based keyframe extraction. The paper reports quantitative gains over random, inverse, and nearest-neighbor baselines on 61 static targets, ablation results on 35 sequential targets, and qualitative physical demonstrations for ASL gestures, shadow puppetry, and animal-motion imitation.

Significance. If the central claim is fully supported, this is an original and engaging contribution: to my knowledge it is the first robotic system that treats projected shadows as a controllable communication channel, and the proposed combination of a learned forward shadow self-model with physics-based refinement is a sensible way to handle the many-to-one projection from hand pose to silhouette. The paper has concrete strengths: the shadow self-model is trained on task-agnostic self-exploration data and frozen during deployment, avoiding circularity in the core imitation result; the analytic-FK-plus-decoder architecture is a reasonable inductive bias; the hybrid optimization pipeline is clearly described; the soft-skinned hand design is motivated by a real light-leakage problem; and the authors release code and supplementary videos. However, the current evidence does not yet fully establish the physical-robot claim: the simulator's visual mesh is the sole source of training data and refinement, yet its fidelity to the real soft-skin shadow is never quantitatively validated, and all quantitative tables report single-run metrics without variance. These issues are load-bearing for the paper's c

major comments (4)
  1. The central claim is that a physical robot can intentionally control its projected shadow. Yet the entire training and optimization pipeline operates in simulation: the shadow self-model is trained on simulated renderings, and the hill-climbing refinement runs in the same simulator. Figure 2 states that the simulator's 'visual mesh approximates the soft skin to reproduce light occlusion effects,' but no quantitative validation of that approximation is provided. There is no comparison of simulator-rendered shadows against real captured shadows for matched joint configurations, and no test of how soft-skin deformation during finger contact changes the silhouette. The physical demonstrations in Figures 4, 7, and S8–S13 are qualitative only. Because the simulator is the sole source of ground truth for both training and refinement, a systematic mismatch between the visual mesh and the real fo
  2. All quantitative results in Tables 1 and 2 are reported as single numbers with no variance, number of seeds, or statistical significance. This is particularly problematic in Table 2, where the central comparisons are close: e.g., 'Ours (with Exp.)' total loss 0.1781 vs 'Ours (with Exp.+ Temp.)' 0.1789. Without multiple random restarts or seeds, it is impossible to tell whether these differences are meaningful or noise. The transition success rates (5/29 vs 25/29) are large, but they depend on thresholds that are introduced in the text without sensitivity analysis. The authors should report mean ± std over at least three optimization seeds (or random initializations) and, where appropriate, a paired test across targets.
  3. The expressive-region ablation is partly circular. Eq. (3) adds weighted Exp. IoU and Exp. CLIP losses to the optimization objective, and Table 2 then shows that 'Ours (with Exp.)' reduces Exp. IoU loss and Exp. CLIP loss relative to 'Ours (Base)'. This reduction is expected by construction, because those exact quantities are being optimized. More importantly, the same table shows that the global reconstruction metrics worsen: Base loss increases from 0.0861 to 0.1141, MAE from 0.1039 to 0.1577, and IoU loss from 0.1950 to 0.2572. The paper claims that expressive-region objectives improve 'motion fidelity and perceptual quality,' but the only evidence is the optimized objective itself and qualitative side-by-side images. The authors should either provide an independent perceptual evaluation (e.g., human ratings) or a held-out perceptual metric, and should report the trade-off against glo
  4. The paper's dynamic-shadow claim rests on the ability to reproduce not just individual silhouettes but temporally coherent motion. The proposed keyframe extraction reduces optimization targets by 63–91.7%, which is a practical advantage, but the paper does not quantitatively measure how well the optimized keyframe sequence preserves the dynamics of the original video. Figure 7B and Figures S8–S13 provide qualitative comparisons, and Table 2 reports transition success rates under a specific binary threshold, but no silhouette-dynamics metric (e.g., motion energy, frame-to-frame silhouette distance, or DTW-based similarity) is reported. Given that one of the three core contributions is dynamic shadow expression, the authors should add a quantitative comparison between the original video's shadow dynamics and the robot-generated sequence, ideally on the physical robot as well as in simulati
minor comments (5)
  1. Typo: 'not-a-knot oundary conditions' should be 'not-a-knot boundary conditions'.
  2. Typo: 'signal-frame optimization' should be 'single-frame optimization'.
  3. The caption says 'Raven flipping wings'; the text elsewhere uses 'flapping'. Please standardize.
  4. The labels 'Independent Per-frame Optimization' and 'Sequential Optimization with Inherited Initialization' are presented as table rows but read as section headers. Consider moving them into a separate column or formatting them as grouping headers for clarity.
  5. The runtime comparison in Figure 6B is useful, but the axes and the number of steps for the 'pure hill-climbing' condition should be stated in the caption to make the comparison self-contained.

Circularity Check

1 steps flagged · score 4.0 of 10

Core imitation pipeline is independent; one ablation metric (Exp. IoU / Exp. CLIP) is the same quantity the optimizer minimizes, so part of the reported improvement is by construction.

  1. fitted input called prediction [Results – Shadow imitation for dynamic video targets, Eq. (3) and Table 2]
    "Lexpr = λ Exp IoU (1 − IoU(M ⊙ I_target, M ⊙ I_hat)) + λ Exp CLIP (1 − g_pred/||g_pred||_2 · g_target/||g_target||_2), (3) ... introducing the expressive-region objective substantially reduced expressive-region IoU loss and expressive-region CLIP loss."

    The Exp. IoU and Exp. CLIP terms in Eq. (3) are exactly the quantities minimized during optimization, using the same masks M. Table 2 then reports 'Exp. IoU' and 'Exp. CLIP' as evidence that adding the expressive-region objective improves performance. An optimizer will almost always reduce its own training objective, so that specific improvement is by construction rather than an independent evaluation of shadow-imitation quality. The paper does include some independent evidence (global losses, transition success rate, qualitative finger tracking), so the circularity is partial and confined to this ablation's headline metric.

full rationale

The central derivation is not circular. The shadow self-model is trained on 9,249,784 task-agnostic (joint configuration, simulated shadow) pairs, frozen during deployment, and optimized against externally supplied target silhouettes from ASL, human hand-shadow puppetry, and raw animal videos. The target images/videos are not derived from the robot's own outputs, and the learned model is not fitted to the test targets. The one load-bearing self-referential element is in the dynamic ablation: Eq. (3) optimizes Exp. IoU and Exp. CLIP losses, and Table 2 reports reductions in those same quantities as a key quantitative improvement, making part of that gain true by construction. The unquantified sim-to-real shadow fidelity (the simulator's visual mesh 'approximates the soft skin to reproduce light occlusion effects') is a genuine correctness/validation gap but is not circularity: the simulator is a separate physical approximation, not a restatement of the paper's conclusions. Self-citations appear only as background (e.g., refs. 18, 34, 35) and are not used to justify the central claim, so they do not raise the score. Overall: the static imitation result is independent, while the dynamic ablation's headline expressive-region metrics are partially fitted-input-called-prediction.

Assumptions & free parameters 10 free parameters · 6 assumptions · 0 invented entities

The central claim depends less on exotic axioms than on domain assumptions about simulation fidelity and metric validity. The paper introduces no new physical entities; its 'shadow self-model' is a learned neural surrogate, not a new phenomenon. The main burden is the assumed match between the simulator's visual mesh and the real soft-skin light occlusion, plus use of CLIP as a perceptual measure for binary silhouettes.

free parameters (10)
  • λ_MAE = 0.06
    Hand-chosen weight for pixel MAE term in Eq. (1); affects all static/dynamic optimization results.
  • λ_IoU = 0.4
    Hand-chosen weight for silhouette overlap term in Eq. (1).
  • λ_CLIP = 0.02
    Hand-chosen weight for CLIP perceptual term in Eq. (1); the same embedding model is also used in evaluation.
  • λ_Exp_IoU = 1.0
    Hand-chosen weight for expressive-region IoU in Eq. (3); drives the reported Exp. IoU improvements.
  • λ_Exp_CLIP = 0.05
    Hand-chosen weight for masked CLIP term in Eq. (3).
  • λ_temp = 0.2
    Hand-chosen weight for temporal smoothness in Eq. (4).
  • Keyframe clustering threshold τ = not reported
    Similarity threshold controlling the number of keyframes per video; directly determines which frames are optimized and the reported 63–92% reductions.
  • Training joint sampling ranges = wrist [-90,90]; MCP thumb [0,90], ring/pinky [0,60], index/middle [-60,0]; other joints approx [-90,90]
    Predefined hand-chosen ranges used to generate the 9.25M self-model training pairs; define the model's coverage of configuration space.
  • Optimization budgets = 2000/3000/1200 iterations; batch 256/500/128; hill-climb 8 runs x 1000 steps, noise SDs 0.25/0.08
    Hand-chosen computational budgets and noise scales; affect success rates and runtime comparison (Fig. 6B).
  • Binarization/cropping thresholds = pixel < 0.5 for crop; intensity 128 for human-hand videos; SAM for animal videos
    Hand-chosen preprocessing thresholds determining what counts as shadow foreground in targets and predictions.
assumptions (6)
  • domain assumption The simulator's visual mesh approximates the real soft-skin geometry/opacity and its deformation under contact.
    Cited in Figure 2 caption and Methods: 'the visual mesh approximates the soft skin to reproduce light occlusion effects.' If this approximation fails, optimized shadow poses will not transfer to hardware. No direct quantitative sim-to-real shadow comparison is provided.
  • domain assumption The learned forward shadow self-model, trained under random joint sampling, remains accurate for the optimized configurations found by gradient descent.
    Test MAE 0.00643 is reported on the same sampling distribution; no evaluation on the optimization trajectory's distribution is given.
  • domain assumption CLIP embeddings provide a useful perceptual similarity signal for binary shadow silhouettes.
    CLIP is trained on natural images/text; using it on thresholded hand/animal silhouettes is an unvalidated transfer. It is both an optimization objective (Eq. 1) and evaluation metric.
  • domain assumption The fixed spotlight, backdrop, and camera setup are stable and match the simulation projection geometry.
    Paper limits itself to controlled illumination; Discussion acknowledges varying conditions remain future work.
  • standard math Standard rigid-body forward kinematics (Rodrigues' formula, URDF) describe the physical hand.
    Eqs. (5)-(7) are textbook kinematics; assumed correct for the 21-DoF hand.
  • domain assumption The 21 joint limits and collision geometry used in simulation match the physical servos and skeleton.
    Motion planning and hill-climbing rely on these; no calibration or tolerance analysis reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Robot Learning to Communicate through Projected Visual Abstractions." pith.science (2026). https://pith.science/paper/NRVYOTRQ

@misc{pith2026260722434,
  author       = {Pith},
  title        = {Pith review of: Robot Learning to Communicate through Projected Visual Abstractions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NRVYOTRQ}},
  note         = {Machine review of arXiv:2607.22434}
}
read the original abstract

Humans routinely communicate through abstractions of their bodies, including shadows, silhouettes, and reflections. Yet robots remain largely confined to expressing themselves through their physical morphology. Enabling robots to communicate through such projected visual abstractions requires reasoning not only about bodily motion but also about how that motion is transformed into an external representation perceived by an observer. Among these abstractions, shadows provide a particularly compelling example because they emerge directly from the robot's embodiment while remaining visually distinct from the body itself. Here, we present a robotic system capable of dynamic shadow expression using a 21-degree-of-freedom dexterous hand with compliant soft skin and a learned shadow self-model. The soft-skinned embodiment reduces light leakage to produce visually continuous silhouettes, while the differentiable self-model learns the mapping between hand configurations and projected shadow appearance through task-agnostic self-exploration. Given a target shadow image or video, the robot optimizes its hand configurations through gradient-based search over 1 the learned self-model and refines the solution through collision-aware simulation to obtain physically feasible motions. For dynamic shadow performance, we further introduce expressive-region objectives, temporal smoothness regularization, and keyframe-based optimization to preserve visually important motion cues while reducing optimization complexity. We demonstrate robotic shadow expression across sign-language gestures, hand-shadow puppetry, and animal motion imitation in both simulation and physical experiments. These results establish a framework for enabling robots to manipulate projected visual abstractions of themselves for communication and visual storytelling.

Figures

Figures reproduced from arXiv: 2607.22434 by the authors.

Figure 4
Figure 4. Shadow imitation for dynamic video targets While single-image shadow imitation evaluates whether our robot can reproduce a desired projected appearance, dynamic shadow targets introduce additional challenges. Beyond matching individual silhouettes, the robot must preserve visually important details, maintain temporally coherent motion, 16 [PITH_FULL_IMAGE:figures/full_fig_p016_4.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 1 canonical work pages

  1. [1]

    T. W. Deacon.The symbolic species: The co-evolution of language and the brain. WW Norton & Company, 1998

    1998

  2. [2]

    Donald.Origins of the modern mind: Three stages in the evolution of culture and cognition

    M. Donald.Origins of the modern mind: Three stages in the evolution of culture and cognition. Harvard university press, 1993

    1993

  3. [3]

    Tomasello.Origins of human communication

    M. Tomasello.Origins of human communication. MIT press, 2010

    2010

  4. [4]

    J. E. Fan, W. A. Bainbridge, R. Chamberlain, J. D. Wammes, Drawing as a versatile cognitive tool. Nature reviews psychology, 2(9), 556–568, 2023

    2023

  5. [5]

    J. E. Fan, R. D. Hawkins, M. Wu, N. D. Goodman, Pragmatic inference and visual abstraction enable contextual flexibility during visual communication. Computational Brain & Behavior, 3(1), 86–101, 2020

    2020

  6. [6]

    Tversky.Mind in motion: How action shapes thought

    B. Tversky.Mind in motion: How action shapes thought. Basic Books, 2019

    2019

  7. [7]

    McNeill.Hand and mind: What gestures reveal about thought

    D. McNeill.Hand and mind: What gestures reveal about thought. University of Chicago press, 1992

    1992

  8. [8]

    Goldin-Meadow, M

    S. Goldin-Meadow, M. W. Alibali, Gesture’s role in speaking, learning, and creating language. Annual review of psychology, 64(1), 257–283, 2013

    2013

Show all 51 references
  1. [9]

    V. I. Stoichita.Short history of the shadow. Reaktion Books, 1997

    1997

  2. [10]

    Casati.Shadows: Unlocking their secrets, from plato to our time

    R. Casati.Shadows: Unlocking their secrets, from plato to our time. Vintage, 2007

    2007

  3. [11]

    Lakoff, M

    G. Lakoff, M. Johnson.Metaphors we live by. University of Chicago press, 2008

    2008

  4. [12]

    Schiff, J

    W. Schiff, J. A. Caviness, J. J. Gibson, Persistent fear responses in rhesus monkeys to the optical stimulus of” looming”. Science, 136(3520), 982–983, 1962

    1962

  5. [13]

    Yilmaz, M

    M. Yilmaz, M. Meister, Rapid innate defensive responses of mice to looming visual stimuli. Current Biology, 23(20), 2011–2015, 2013. 32

    2011

  6. [14]

    Oliva, V

    D. Oliva, V. Medan, D. Tomsic, Escape behavior and neuronal responses to looming stimuli in the crab Chasmagnathus granulatus (Decapoda: Grapsidae). Journal of Experimental Biology, 210(5), 865–880, 2007

    2007

  7. [15]

    Schleidt, M

    W. Schleidt, M. D. Shalter, H. Moura-Neto, The hawk/goose story: the classical ethological experiments of Lorenz and Tinbergen, revisited. Journal of comparative psychology, 125(2), 121, 2011

    2011

  8. [16]

    Bongard, V

    J. Bongard, V. Zykov, H. Lipson, Resilient machines through continuous self-modeling. Sci- ence, 314(5802), 1118–1121, 2006

    2006

  9. [17]

    Kwiatkowski, H

    R. Kwiatkowski, H. Lipson, Task-agnostic self-modeling machines. Science Robotics, 4(26), eaau9354, 2019

    2019

  10. [18]

    B. Chen, R. Kwiatkowski, C. Vondrick, H. Lipson, Fully body visual self-modeling of robot morphologies. Science Robotics, 7(68), eabn1944, 2022

    2022

  11. [19]

    Y. Hu, B. Chen, H. Lipson, Egocentric visual self-modeling for autonomous robot dynamics prediction and adaptation. npj Robotics, 3(1), 14, 2025

    2025

  12. [20]

    T. D. Kulkarni, W. F. Whitney, P. Kohli, J. Tenenbaum, Deep convolutional inverse graphics network. Advances in neural information processing systems, 28, 2015

    2015

  13. [21]

    D. Ha, J. Schmidhuber, Recurrent world models facilitate policy evolution. Advances in neural information processing systems, 31, 2018

    2018

  14. [22]

    Hafner, T

    D. Hafner, T. Lillicrap, J. Ba, M. Norouzi. Dream to Control: Learning Behaviors by Latent Imagination. InInternational Conference on Learning Representations, 2020. available at https://openreview.net/forum?id=S1lOTC4tDS

    2020

  15. [23]

    J. G. Trafton, N. L. Cassimatis, M. D. Bugajska, D. P. Brock, F. E. Mintz, A. C. Schultz, Enabling effective human-robot interaction using perspective-taking in robots. IEEE Transactions on Systems, Man, and Cybernetics-Part A: Systems and Humans, 35(4), 460–470, 2005. 33

    2005

  16. [24]

    Fischer, Y

    T. Fischer, Y. Demiris. Markerless perspective taking for humanoid robots in unconstrained environments. In2016 IEEE International conference on robotics and automation (ICRA), pages 3309–3316. IEEE, 2016

    2016

  17. [25]

    B. Chen, Y. Hu, R. Kwiatkowski, S. Song, H. Lipson. Visual perspective taking for opponent behavior modeling. In2021 IEEE International Conference on Robotics and Automation (ICRA), pages 13678–13685. IEEE, 2021

    2021

  18. [26]

    B. Chen, C. Vondrick, H. Lipson, Visual behavior modelling for robotic theory of mind. Scientific Reports, 11(1), 424, 2021

    2021

  19. [27]

    Z. Ji, L. Zhang, P. Sajda, B. Chen. Enabling multi-robot collaboration from single-human guidance. In2025 IEEE International Conference on Robotics and Automation (ICRA), pages 4272–4279. IEEE, 2025

    2025

  20. [28]

    Aguilar, H

    C. Aguilar, H. Lipson. A robotic system for interpreting images into painted artwork. In International conference on generative art, volume 11, 2008

    2008

  21. [29]

    Y. Sun, Y. Xu. A calligraphy robot—Callibot: Design, analysis and applications. In2013 IEEE international conference on robotics and biomimetics (ROBIO), pages 185–190. IEEE, 2013

    2013

  22. [30]

    S. Song, J. Kim, K. Yamane. Development of a bipedal robot that walks like an animation character. In2015 IEEE International Conference on Robotics and Automation (ICRA), pages 3596–3602. IEEE, 2015

    2015

  23. [31]

    M. A. Hopkins, G. Wiedebach, K. Cesare, J. Bishop, E. Knoop, M. B ¨acher, Interactive design of stylized walking gaits for robotic characters. ACM Transactions on Graphics (ToG), 43(4), 1–16, 2024

    2024

  24. [32]

    M¨ uller, E

    D. M¨ uller, E. Knoop, D. Mylonopoulos, A. Serifi, M. A. Hopkins, R. Grandia, M. B ¨acher, Olaf: Bringing an Animated Character to Life in the Physical World. arXiv preprint arXiv:2512.16705, 2025

    2025

  25. [33]

    B. Chen, Y. Hu, L. Li, S. Cummings, H. Lipson. Smile like you mean it: Driving animatronic robotic face with learned models. In2021 IEEE International Conference on Robotics and Automation (ICRA), pages 2739–2746. IEEE, 2021. 34

    2021

  26. [34]

    Y. Hu, B. Chen, J. Lin, Y. Wang, Y. Wang, C. Mehlman, H. Lipson, Human-robot facial coexpression. Science Robotics, 9(88), eadi4724, 2024

    2024

  27. [35]

    Y. Hu, J. Lin, J. A. Goldfeder, P. M. Wyder, Y. Cao, S. Tian, Y. Wang, J. Wang, M. Wang, J. Zeng, C. Mehlman, Y. Wang, D. Zeng, B. Chen, H. Lipson, Learning realistic lip motions for humanoid face robots. Science Robotics, 11(110), eadx3017, 2026, doi:10.1126/scirobotics.adx30...

    2026 doi

  28. [36]

    Fern ´andez-Navarro, O

    V. Fern ´andez-Navarro, O. Spaey, D. Garate, Reevaluating Hand Stencil Phenomena in Cave Art: A Step Forward towards the Characterization of Symbolic Patterns during the Upper Palaeolithic in Europe. Cambridge Archaeological Journal, 1, 21, 2026

    2026

  29. [37]

    Leroi-Gourhan.The dawn of European art: an introduction to Palaeolithic cave painting

    A. Leroi-Gourhan.The dawn of European art: an introduction to Palaeolithic cave painting. CUP Archive, 1982

    1982

  30. [38]

    Currell.Shadow puppets and shadow play

    D. Currell.Shadow puppets and shadow play. Crowood, 2015

    2015

  31. [39]

    F. P. Chen, Shadow theaters of the world. Asian Folklore Studies, 25–64, 2003

    2003

  32. [40]

    F.-P. L. Chen.Chinese shadow theatre: History, popular religion, and women warriors. McGill-Queen’s Press-MQUP, 2007

    2007

  33. [41]

    Zhang, Y

    H. Zhang, Y. Song, Z. Chen, J. Cai, K. Lu. Chinese shadow puppetry with an interactive interface using the Kinect sensor. InEuropean Conference on Computer Vision, pages 352–

  34. [42]

    Almoznino, Y

    A. Almoznino, Y. Pinas.The art of hand shadows. Courier Corporation, 2002

    2002

  35. [43]

    The Shadow Only

    N. Forgione, “The Shadow Only”: Shadow and Silhouette in Late Nineteenth-Century Paris. The Art Bulletin, 81(3), 490–512, 1999

    1999

  36. [44]

    Tsuchiya, T

    M. Tsuchiya, T. Itoh, M. Neff, Y. Liu. A Virtual and Interactive Light-Art-Like Representation of Human Silhouette. In2019 International Conference on Cyberworlds (CW), pages 403–406. IEEE, 2019. 35

    2019

  37. [45]

    N. J. Mitra, M. Pauly, Shadow art. ACM Transactions on Graphics, 28(5), 156–1, 2009

    2009

  38. [46]

    Bermano, I

    A. Bermano, I. Baran, M. Alexa, W. Matusk. Shadowpix: Multiple images from self shadowing. InComputer Graphics Forum, volume 31, pages 593–602. Wiley Online Library, 2012

    2012

  39. [47]

    J. Won, J. Lee, Shadow theatre: discovering human motion from a sequence of silhouettes. ACM Transactions on Graphics (TOG), 35(4), 1–12, 2016

    2016

  40. [48]

    H. Xu, Y. Wang, N. J. Mitra, S. Liu, P.-A. Heng, C.-W. Fu, Hand-shadow poser. arXiv preprint arXiv:2505.07012, 2025

    2025 arXiv

  41. [49]

    Zhang, H

    Y. Zhang, H. Jiang, Y. Miura, C. D. Manning, C. P. Langlotz. Contrastive learning of med- ical visual representations from paired images and text. InMachine learning for healthcare conference, pages 2–25. PMLR, 2022

    2022

  42. [50]

    Radford, J

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark, others. Learning transferable visual models from natural language super- vision. InInternational conference on machine learning, pages 8748–8763. PmLR, 2021

    2021

  43. [51]

    Kirillov, E

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y. Lo, others. Segment anything. InProceedings of the IEEE/CVF international conference on computer vision, pages 4015–4026, 2023. Acknowledgments: We thank Jacob Lee for t...

    2023

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.