Pith. sign in

REVIEW 3 major objections 5 minor 68 references

The FlEye camera: Sampling the joint distribution of natural scenes and motion

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read The FlEye camera records natural scenes with a fly's optics and high signal-to-noise ratio, showing that optimal yaw-velocity estimation crosses from correlator-like to gradient-like behavior and that pitch-direction gradients act as…

desk verdict FlEye is a solid instrument paper; the 3D motion-estimator claim is undersampled and needs error bars before it can support the connectivity prediction. read the letter →

arxiv 2412.21081 v1 pith:2FWW6AHJ submitted 2024-12-30 q-bio.NC cond-mat.dis-nn

Charles J. Edelson , Paul Smith , Sima Setayeshgar , William Bialek , Rob R. de Ruyter van Steveninck This is my paper · ORCID
classification q-bio.NCcond-mat.dis-nn
keywords flyvisualsystemmotionestimationnaturalscenestatisticscompoundeyeopticsoptimalgradientestimatorcorrelatormodelinertialmeasurementunit
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 reports a portable camera, the FlEye, that sees the way a blowfly's compound eye does: a hexagonal lattice of photodiodes with the fly's inter-receptor spacing, Gaussian point-spread functions of the right width, matched spectral sensitivity, and 1000 Hz sampling, with an inertial unit that records the camera's own rotations at 2000 Hz. Because the photodiodes operate at an effective photon rate above $10^8$ s$^{-1}$ while fly photoreceptors saturate near $10^6$ s$^{-1}$, the movies are near-ground-truth light intensities paired with true motion trajectories. Using a 45-minute walk through the woods as a sample of the joint distribution of natural scenes and motion, the authors build optimal local estimators of yaw velocity from derivatives of log intensity. The estimators reproduce the predicted crossover from correlator-like behavior at small derivatives to gradient-like behavior at large derivatives, and the larger dataset reveals that large gradients in the pitch direction, together with pitch movement, act as noise for yaw estimation. If this is right, yaw-sensitive motion neurons in the fly should receive input from photoreceptors separated along the pitch axis, a prediction now checkable in the wiring diagram.

What carries the argument

The central object is the FlEye camera itself: 187 silicon PIN photodiodes in a hexagonal lattice with the blowfly's $1.57^\circ$ inter-receptor spacing, masked to circular apertures and placed behind a holographic diffuser and a two-lens system that reproduces the compound eye's Gaussian point-spread function of width about $0.51^\circ$, plus a cyan filter matching the photoreceptor spectral peak near 490 nm. Its effective photon-counting rate exceeds $10^8$ s$^{-1}$, orders of magnitude above the fly's saturation near $10^6$ s$^{-1}$, so the recorded intensities are close to ground truth and fly-like noise can be added later. An inertial measurement unit is synchronized to the photodiodes at the millisecond level, so each frame is paired with pitch, roll, and yaw velocities. The analytic machinery is the conditional-mean estimator $\hat v_\phi = E[v_\phi | \partial_t \ln I, \partial_\phi \ln I, \partial_\theta \ln I]$, evaluated by Monte Carlo binning of the measured joint distribution; slicing this estimator at fixed $\partial_\theta \ln I$ is what exposes both the gradient-to-correlator crossover and the cross-axis noise effect. In that crossover, the gradient estimator, which forms velocity as the ratio of temporal to spatial derivatives, is the high-signal-to-noise limit, while the correlator, which forms velocity from their product, is the low-signal-to-noise limit.

What would settle it

Compute the same conditional-mean estimators from image-and-motion data recorded on a freely flying insect or on a platform driven with fly-like accelerations; if the gradient-to-correlator crossover or the pitch-gradient noise effect changes qualitatively, the naturalistic-sample assumption is refuted. Separately, inspect the connectome of yaw-sensitive lobula-plate neurons: if their inputs do not include photoreceptors separated along the pitch axis, the paper's central neural prediction is refuted.

Watch

Extended reading notes

Core claim

The central discovery is that natural movie-and-motion data, collected with a fly-matched camera, support the optimal-estimation picture of fly motion vision and expose a new cross-axis interaction. When yaw velocity is estimated from the local temporal and azimuthal derivatives of log intensity, the conditional-mean estimator $\hat v_\phi = E[v_\phi | \partial_t \ln I, \partial_\phi \ln I]$ has contours that are correlator-like hyperbolae at small derivatives and gradient-like straight lines at large derivatives, with the crossover at derivative magnitudes that occur during an ordinary walk. Conditioning also on the polar derivative $\partial_\theta \ln I$ shows that large $|\partial_\theta \ln I|$, combined with pitch motion, expands the ill-defined central region and makes the yaw-velocity contours noisier, which the authors read as orthogonal gradients acting as noise for yaw estimation. They draw the concrete consequence that yaw-sensitive motion neurons should receive inputs from photoreceptors separated along the pitch axis.

Load-bearing premise

The load-bearing premise is that a 45-minute walk through the woods by a person samples the joint distribution of visual inputs and motions that flies encounter in nature; the paper itself notes (Section IV B) that human walking does not reach the enthusiasm of fly flight, so if fly motion statistics differ substantially, the measured estimator structure, including the pitch-gradient noise effect, may not transfer to the fly.

Editorial extensions

If this is right

  • The gradient-to-correlator crossover in optimal local motion estimators is visible in natural data and occurs at derivative values that are common on a real walk, so the predicted systematic errors of motion estimation matter for real vision, not only for toy stimuli.
  • Adding simulated photoreceptor noise to the camera's near-noiseless records expands the correlator-like region, qualitatively matching experiments on motion-sensitive neurons; estimation errors are thus consistent with optimal inference in noise rather than a biological flaw.
  • Yaw-velocity estimates are degraded by large pitch-direction gradients combined with pitch motion, so nominally orthogonal visual signals are not irrelevant to a given velocity estimate.
  • Yaw-sensitive lobula-plate neurons should receive input from photoreceptors separated along the pitch axis, a prediction that can be checked directly against the recently mapped connectome.
  • The instrument enables a calibrated database of natural movies paired with ground-truth motion, providing a resource for quantitative tests of adaptation, wide-field motion integration, and coding efficiency in the fly visual system.

Reading between the lines

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

  • The authors do not test whether the cross-axis interaction is reciprocal, but the same estimator construction could check whether pitch-velocity estimates are analogously degraded by large azimuthal gradients, which would strengthen the claim that orthogonal derivatives act as noise.
  • Because the training data come from human walking, which the paper itself flags as gentler than fly flight, a natural extension is to record from a camera carried by a free-flying insect or a fast robot; the prediction is that the sizes of the correlator region and the pitch-noise effect shift with motion statistics while the qualitative structure remains.
  • The same Monte Carlo conditional-mean approach could be applied to nonlocal features such as pairs or triplets of spatial derivatives, asking how much information local estimators leave on the table compared with wide-field integration.
  • Adding calibrated photon shot noise to the stored movies would turn the database into a tunable signal-to-noise dial, allowing a systematic map of the gradient-to-correlator transition and its pitch-axis modulation across light levels.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper reports the design, construction, and calibration of the FlEye camera, a portable fly-eye-like imaging system with 187 photodiodes, fly-matched spectral sensitivity and point-spread functions, and a synchronized IMU for motion capture. After characterizing the instrument (PSFs, noise power spectra, transfer functions, effective Poisson rate, and IMU-optical alignment), the authors use a 45-minute walk through a nature preserve to construct optimal local estimators of yaw velocity from temporal and spatial derivatives of log intensity. The two-dimensional estimator reproduces the previously reported gradient-to-correlator crossover, and a three-dimensional extension conditioning also on the polar derivative is used to argue that large polar gradients act as noise for yaw estimation, leading to a prediction about yaw-sensitive connectivity in the fly visual system.

Significance. If the instrument claims stand, this is a valuable contribution: the camera characterization is unusually complete, including measured PSF covariance matrices and skewness, noise power spectra, contrast transfer functions, an effective Poisson rate exceeding 10^8 Hz, and a quantitative IMU-optical alignment check with 0.54 deg/s mean difference. The paper also provides reproducible code and data-processing tools, and it makes a concrete, falsifiable biological prediction. The two-dimensional estimator results are a clear and useful illustration of the instrument's potential. However, the new three-dimensional claim in Section IV C and the resulting Section V prediction currently rest on a binned conditional mean without uncertainty quantification, so the statistical strength of the central new result is not yet established.

major comments (3)
  1. [IV C, Eq. (13), Fig. 8] The new biological claim rests on the three-dimensional conditional mean in Eq. (13), but Fig. 8 is presented without error bars, confidence intervals, or per-bin sample counts. The text states that conditional means are computed using N=100 equal-sized bins along each axis; if this same binning is used for the three-dimensional estimator, the ~1.35e6 samples available after 500 Hz downsampling give on average about one sample per bin. Because the derivative distributions have nearly exponential tails (Fig. 6b), the high-|∂θ ln I| slices that display the reported 'cone' and 'fuzziness' must contain very few samples. The authors should show, via bootstrap or jackknife confidence intervals, adaptive or coarser binning, or a regularized estimator, that the cone and the loss of definition survive sampling noise; otherwise the qualitative reading of Fig. 8 is not distinguishable from Monte Carlo fluctuations. If a different binning was used for the three-dimensional case, that binning should be stated explicitly.
  2. [IV C, text after Fig. 8] The statement that 'large gradients in the polar or pitch direction, in combination with pitch movement, act as noise for estimating yaw velocity' is not directly supported by Eq. (13), which conditions on ∂θ ln I but not on any IMU measure of pitch velocity. The estimator in Eq. (13) cannot by itself separate the effect of large polar gradients from the effect of pitch motion, even if the two are correlated in natural data. To support the 'in combination with pitch movement' claim, the authors should include pitch velocity in the conditioning set, or analyze the interaction between ∂θ ln I and measured pitch velocity explicitly. The current analysis only establishes, at best, a dependence of the conditional mean on ∂θ ln I, not the proposed mechanism.
  3. [IV B / V] The paper acknowledges in Section IV B that 'human walking does not reach the enthusiasm of fly flight,' yet the Section V prediction about yaw-sensitive connectivity is stated without qualification. If fly flight has substantially different joint statistics—for example, higher angular velocities, saccadic head and body movements, or different correlations between pitch and yaw—the orthogonal-gradient interaction observed in the human-walk data may not transfer to the fly. The authors should either temper the connectivity prediction or provide evidence that the estimator structure, especially the ∂θ ln I dependence, is robust across motion statistics, for instance by reweighting the data to better match fly-like movement distributions or by recording under additional natural conditions.
minor comments (5)
  1. [Introduction] There is a typo in 'large monpoloar cells (LMCs)'; it should read 'large monopolar cells.'
  2. [Section III B / Eq. (5)] In the text preceding Eq. (5), 'Possion rate' should be 'Poisson rate.'
  3. [Section II B] The sentence 'the photodiodes were masked to a 1 mm circular active area to remove any affects their square profile had on the PSF' should use 'effects' rather than 'affects.'
  4. [Appendix C] The heading 'Performance measurements' is misspelled as 'meausurements,' and later in the same appendix 'esitmate' should be 'estimate.'
  5. [Fig. 4 caption] The caption says 'saturare at λeff ∼ 10^6 Hz'; this should be 'saturate.'

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the estimator is an empirical conditional mean computed from the recorded data, and the neural-connectivity prediction is an extrapolation beyond the fitted surface.

full rationale

The paper's derivation chain is self-contained. The optimal local motion estimator is defined directly as the conditional mean E[v_phi | derivatives] (Eqs. 11 and 13) and computed by Monte Carlo averaging over the recorded FlEye data; therefore the appearance of gradient-like and correlator-like regions in Fig. 7 is an empirical description of the joint statistics, not a quantity fitted elsewhere and then renamed as a prediction. The new three-dimensional result in Fig. 8 is likewise a direct conditional average, and the Section V statement that yaw-sensitive neurons should receive inputs from pitch-separated photoreceptors is an extrapolation beyond the fitted surface, not a restatement of the regression input. Self-citations [44], [56], and [62] are present, but [44] is used only for comparison with an earlier data set, [56] is a theoretical result motivating the shape of the estimator, and [62] is a peripheral neural hint; removing any of them would not collapse the central claim. The statistical concern about sparse high-|d_theta ln I| slices in Fig. 8 is a robustness and sampling issue for the empirical estimate, not a circularity of the derivation. No step in the paper reduces by construction to its own input.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The account is largely empirical: the camera is a physical instrument, and the estimator is a data-conditional average. The main hidden costs are the per-photodiode calibration fits and the assumption that one human walk captures the fly's natural input distribution.

free parameters (3)
  • Per-photodiode one-knot quadratic spline coefficients = theta0-theta4 and knot k1 per photodiode
    Calibration from raw 20-bit output to radiance, fit by fminsearch to 29 radiance levels (Appendix D); the estimator uses these calibrated intensities.
  • Estimator bin count N = 100 bins per derivative axis
    Binning choice for the empirical conditional mean in Section IV C; no sensitivity analysis is provided.
  • Preprocessing parameters = 2-frame moving average; downsampling to 500 Hz
    Section IV B; these choices affect derivative estimates and the signal-to-noise ratio of the estimator surface.
assumptions (5)
  • domain assumption Rigid motion model I(phi,t) = I0(phi - vt)
    Introduced in Section IV A Eq (6); assumes the world moves rigidly with yaw velocity v and ignores parallax, translation, and scene depth.
  • standard math Conditional mean minimizes mean-square error
    Eqs (9)-(10); a standard result in estimation theory, used to define the optimal estimator.
  • domain assumption Local spatiotemporal derivatives are sufficient statistics for local velocity estimation
    Eq (11) conditions only on d_t ln I and d_phi ln I (and later d_theta ln I); the camera data contain richer information, so the estimator is optimal only within this restricted information set.
  • domain assumption A single 45-minute human walk represents the fly's natural environment
    Section IV B uses one walk through Porter West Nature Preserve; the authors acknowledge 'human walking does not reach the enthusiasm of fly flight', so the sampled distribution may not match fly behavior.
  • domain assumption Noise is photon shot noise, Ns(omega) = |T(omega)|^2 / lambda_eff
    Eq (5) defines the effective Poisson rate under this assumption; the camera's noise may have additional electronic components.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The FlEye camera: Sampling the joint distribution of natural scenes and motion." pith.science (2026). https://pith.science/paper/2FWW6AHJ

@misc{pith2026241221081,
  author       = {Pith},
  title        = {Pith review of: The FlEye camera: Sampling the joint distribution of natural scenes and motion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2FWW6AHJ}},
  note         = {Machine review of arXiv:2412.21081}
}
read the original abstract

To make efficient use of limited physical resources, the brain must match its coding and computational strategies to the statistical structure of input signals. An attractive testing ground for these principles is the problem of motion estimation in the fly visual system: we understand the optics of the compound eye, have a quantitative description of input signals and noise from the retina, and can record from output neurons that encode estimates of different velocity components. Furthermore, recent work provides a nearly complete wiring diagram of the intervening circuitry. What is missing is a characterization of the visual signals and motions that flies encounter in a natural context. We attack this directly with the development of a specialized camera that matches the high temporal resolution, optical properties, and spectral sensitivity of the fly's eye; inertial motion sensors provide ground truth about rotations and translations through the world. We describe the design, construction, and performance characteristics of this FlEye camera. To illustrate the opportunities created by this instrument we use data on movies and motion to construct optimal local motion estimators that can be compared with the responses of the fly's motion sensitive neurons.

Figures

Figures reproduced from arXiv: 2412.21081 by the authors.

Figure 1
Figure 1. FIG. 1. The FlEye camera. (a) Completed and assem [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. FlEye camera optics. (a) Schematic showing the im [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. Optically measured average photodiode point spread [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: FIG. 4. Signal and noise characteristics of the photodiodes. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5. Optical vs direct mechanical measures of yaw veloc [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6. Statistics of the natural scenes recorded on a 45 [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7. Comparison of the optimal estimators of (yaw) an [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: FIG. 8. The optimal estimator of yaw velocity from Eq [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 9
Figure 9. Figure 9: FIG. 9. Raw write data format for the FlEye camera. Individ [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]
Figure 10
Figure 10. Figure 10: FIG. 10. Photodiode to radiance response curves for five [PITH_FULL_IMAGE:figures/full_fig_p013_10.png]
Figure 11
Figure 11. Figure 11: FIG. 11. AIC distributions for each of the models listed in Ta [PITH_FULL_IMAGE:figures/full_fig_p014_11.png]
Figure 12
Figure 12. Figure 12: FIG. 12. Quadratic spline losses across the SSR loss (a), [PITH_FULL_IMAGE:figures/full_fig_p014_12.png]
Figure 13
Figure 13. Figure 13: FIG. 13. Fitted photodiode response curve for the central [PITH_FULL_IMAGE:figures/full_fig_p014_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

68 extracted references · 50 canonical work pages

  1. [1]

    Individual display LEDs had a square pro- file with an edge length of 5 mm, giving the full display a height and width of 24 .2 cm

    Optical Measurements The photodiode PSF was measured by recording the camera’s responses to a 48 × 48 high intensity LED dis- play with a 500 Hz rate and 12-bit pulse width modula- tion (PWM). Individual display LEDs had a square pro- file with an edge length of 5 mm, giving the full display a height and width of 24 .2 cm. The photodiode PSF and effective...

    work page

  2. [2]

    In particular, we wanted the camera’s SNR to exceed that of the fly large monopolar cell (LMC), a secondary visual neuron that combines the responses of six photoreceptors

    Signals and noise One of the major design goals of the camera was for it to have a SNR significantly higher than the fly early vi- sual system. In particular, we wanted the camera’s SNR to exceed that of the fly large monopolar cell (LMC), a secondary visual neuron that combines the responses of six photoreceptors. We quantify the camera’s SNR by looking ...

    work page

  3. [3]

    First, a simple rotation test of the FlEye camera on a high precision turntable

    Measurements of motion The gyro-optical properties of the camera were mea- sured in two experiments. First, a simple rotation test of the FlEye camera on a high precision turntable. The FlEye camera was centered on the table along one of its main rotational axes. The table was then rotated through a half rotation. The IMU’s response was integrated along t...

    work page

  4. [4]

    von Helmholtz, Treatise on Physiological Optics (Op- tical Society of America, Washington DC, 1925) Trans- lated from the 3rd German Edition (1910), by J

    H. von Helmholtz, Treatise on Physiological Optics (Op- tical Society of America, Washington DC, 1925) Trans- lated from the 3rd German Edition (1910), by J. P. C. Southall

    work page 1910

  5. [5]

    C. E. Shannon, The Bell System Technical Journal 27, 379 (1948)

    1948

  6. [6]

    Wiener, Extrapolation, Interpolation, and Smoothing of Stationary Time Series (John Wiley and Sons, New York, 1949)

    N. Wiener, Extrapolation, Interpolation, and Smoothing of Stationary Time Series (John Wiley and Sons, New York, 1949)

    work page 1949

  7. [7]

    J. L. Lawson and G. E. Uhlenbeck, Threshold Signals (McGraw-Hill, New York, 1950) Massachusetts Institute of Technology Radiation Laboratory Series, Volume 24

    work page 1950

  8. [8]

    Grassberger and J.-P

    P. Grassberger and J.-P. Nadal, eds., From Statistical Physics to Statistical Inference and Back (Kluwer, Dor- drecht, 1994)

    work page 1994

Show all 68 references
  1. [9]

    H. B. Barlow, in National Physical Laboratory Sympo- sium on the Mechanization of Thought Processes, Volume 2 (HM Stationery Office, London, 1959) pp. 537–574

    1959

  2. [10]

    Bialek, Biophysics: Searching for Principles (Prince- ton Univerity Press, Princeton NJ, 2012)

    W. Bialek, Biophysics: Searching for Principles (Prince- ton Univerity Press, Princeton NJ, 2012)

    2012

  3. [11]

    Bialek, in Les Houches Summer School Lecture Notes: Theoretical Biological Physics 2023, edited by A.-F

    W. Bialek, in Les Houches Summer School Lecture Notes: Theoretical Biological Physics 2023, edited by A.-F. Bit- bol, T. Mora, I. Nemenman, and A. M. Walczak (SciPost Physics Lecture Notes, 2024)

    2023

  4. [12]

    S. B. Laughlin, Zeitschrift f¨ ur Naturforschung 36c, 910 (1981)

    1981

  5. [13]

    J. H. van Hateren, Nature 360, 68 (1992)

    1992

  6. [14]

    Rieke, D

    F. Rieke, D. Warland, R. de Ruyter van Steveninck, and W. Bialek, Spikes: Exploring the Neural Code (MIT Press, Cambridge MA, 1997)

    1997

  7. [15]

    Brenner, W

    N. Brenner, W. Bialek, and R. de Ruyter van Steveninck, Neuron 26, 695 (2000)

    2000

  8. [16]

    D. A. Clark and J. E. Fitzgerald, Annual Review of Vi- sion Science 10, 23 (2024)

    2024

  9. [17]

    Zbikowski, IEEE Instrumentation and Measurement Magazine 7, 19 (2004)

    R. Zbikowski, IEEE Instrumentation and Measurement Magazine 7, 19 (2004)

    2004

  10. [18]

    Dickson, A

    W. Dickson, A. Straw, C. Poelma, and M. Dickinson, in 44th AIAA Aerospace Sciences Meeting and Exhibit (2006) p. 34

    2006

  11. [19]

    G. K. Taylor and H. G. Krapp, Advances in Insect Phys- iology 34, 231 (2007)

    2007

  12. [20]

    J. C. Tuthill and R. I. Wilson, Current Biology26, R1022 (2016)

    2016

  13. [21]

    Lehmann and J

    F.-O. Lehmann and J. Bartussek, Journal of Compara- tive Physiology A 203, 1 (2017)

    2017

  14. [22]

    B. H. Dickerson, Proceedings of the Royal Society B 287, 20201774 (2020)

    2020

  15. [23]

    W. B. Dickson, A. D. Straw, and M. H. Dickinson, AIAA Journal 46, 2150 (2008)

    2008

  16. [24]

    S. C. Whitehead, S. Leone, T. Lindsay, M. R. Meiselman, N. J. Cowan, M. H. Dickinson, N. Yapici, D. L. Stern, T. Shirangi, and I. Cohen, Science Advances 8, eabo7461 (2022)

    2022

  17. [25]

    de Ruyter van Steveninck and S

    R. de Ruyter van Steveninck and S. B. Laughlin, Inter- national Journal of Neural Systems 7, 437 (1996)

    1996

  18. [26]

    D. Cao, A. J. Zele, and J. Pokorny, Vision Research 47, 1060 (2007)

    2007

  19. [27]

    D. G. Stavenga, Journal of Comparative Physiology A 189, 189 (2003)

    2003

  20. [28]

    H. B. Barlow, The Journal of Experimental Biology 29, 667 (1952)

    1952

  21. [29]

    D. G. Stavenga, Journal of Comparative Physiology A 189, 1 (2003)

    2003

  22. [30]

    D. G. Stavenga, Journal of Comparative Physiology A 190, 115 (2004)

    2004

  23. [31]

    P. G. Lillywhite, Journal of Comparative Physiology A 122, 189 (1977)

    1977

  24. [32]

    R. R. de Ruyter van Steveninck and S. B. Laughlin, Na- ture 379, 642 (1996)

    1996

  25. [33]

    Pierantoni, Cell and Tissue Research 171, 101 (1976)

    R. Pierantoni, Cell and Tissue Research 171, 101 (1976)

    1976

  26. [34]

    Hausen, in Photoreception and Vision in Inverte- brates, edited by M

    K. Hausen, in Photoreception and Vision in Inverte- brates, edited by M. A. Ali (Springer, 1984) pp. 523–559

    1984

  27. [35]

    H. G. Krapp, Navigation 55, 147 (2008)

    2008

  28. [36]

    S. J. Huston and H. G. Krapp, PLoS Biology 6, e173 (2008)

    2008

  29. [37]

    T. W. Ullrich, R. Kern, and M. Egelhaaf, Biology Open 4, 13 (2014)

    2014

  30. [38]

    R. D. Williams, T. L. Massey, and M. M. Maharbiz, in 2018 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC) (IEEE, 2018) pp. 1685–1688

    2018

  31. [39]

    H. Wei, H. Y. Kyung, P. J. Kim, and C. Desplan, Journal of Comparative Physiology A 206, 139 (2020)

    2020

  32. [40]

    S Dorkenwald, A

    A. S Dorkenwald, A. Matsliah, A. R. Sterling, P. Schlegel, S.-C. Yu, C. E. McKellar, A. Lin, M. Costa, K. Eich- ler, Y. Yin, W. Silversmith, C. Schneider-Mizell, C. S. Jordan, D. Brittain, A. Halageri, K. Kuehner, O. Oge- dengbe, R. Morey, J. Gager, K. Kruk, E. Perlman, R. Yan...

    2024

  33. [41]

    de Ruyter van Steveninck and W

    R. de Ruyter van Steveninck and W. Bialek, in Mod- els of Neural Networks IV: Early Vision and Attention (Springer–Verlag, 2002) pp. 313–365

    2002

  34. [42]

    Nemenman, G

    I. Nemenman, G. D. Lewen, W. Bialek, and R. R. de Ruyter van Steveninck, PLoS Computational Biology 4, e1000025 (2008)

    2008

  35. [43]

    M. S. Drews, A. Leonhardt, N. Pirogova, F. G. Richter, A. Schuetzenberger, L. Braun, E. Serbe, and A. Borst, Current Biology 30, 209 (2020)

    2020

  36. [44]

    A. L. Fairhall, G. D. Lewen, W. Bialek, and R. de Ruyter van Steveninck, Nature 412, 787 (2001)

    2001

  37. [45]

    Bialek, F

    W. Bialek, F. Rieke, R. R. de Ruyter van Steven inck, and D. Warland, Science 252, 1854 (1991)

    1991

  38. [46]

    de Ruyter van Steveninck and W

    R. de Ruyter van Steveninck and W. Bialek, Philosophi- cal Transactions of the Royal Society of London348, 321 (1995)

    1995

  39. [47]

    S. R. Sinha, W. Bialek, and R. R. de Ruyter van Steveninck, Physical Review Letters 126, 018101 (2021)

    2021

  40. [48]

    Merrifield and L

    S. Merrifield and L. Wilkinson, SimpleSDHC, https:// github.com/ibm2030/SimpleSDHC (2000)

    2000

  41. [49]

    Beersma, D

    D. Beersma, D. Stavenga, and J. Kuiper, Journal of com- parative physiology 119, 207 (1977)

    1977

  42. [50]

    J. G. J. Smakman, J. H. van Hateren, and D. G. Stavenga, Journal of Comparative Physiology A155, 239 (1984)

    1984

  43. [51]

    C. J. Edelson, MCRayTracing, https://github.com/ edelsonc/MCRayTracing (2024)

    2024

  44. [52]

    Limb and J

    J. Limb and J. Murphy, Computer Graphics and Image 16 Processing 4, 311 (1975)

    1975

  45. [53]

    Reichardt, in Symposium on Principles of Sensory Communication 1959 (MIT Press, Cambridge MA, 1961) pp

    W. Reichardt, in Symposium on Principles of Sensory Communication 1959 (MIT Press, Cambridge MA, 1961) pp. 303–317

    1959

  46. [54]

    E. H. Adelson and J. R. Bergen, Journal of the Optical Society of America A 2, 284 (1985)

    1985

  47. [55]

    Reichardt and T

    W. Reichardt and T. Poggio, Quarterly Reviews of Bio- physics 9, 311 (1976)

    1976

  48. [56]

    Egelhaaf, A

    M. Egelhaaf, A. Borst, and W. Reichardt, Journal of the Optical Society of America A 6, 1070 (1989)

    1989

  49. [57]

    de Ruyter van Steveninck, W

    R. de Ruyter van Steveninck, W. Bialek, M. Potters, and R. Carlson, in Proceedings of IEEE International Con- ference on Systems, Man and Cybernetics , Vol. 1 (IEEE,

  50. [58]

    J. P. Van Santen and G. Sperling, Journal of the Optical Society of America A 1, 451 (1984)

    1984

  51. [59]

    Potters and W

    M. Potters and W. Bialek, Journal de Physique I 4, 1755 (1994)

    1994

  52. [60]

    Weiss, E

    Y. Weiss, E. P. Simoncelli, and E. H. Adelson, Nature Neuroscience 5, 598 (2002)

    2002

  53. [61]

    A. A. Stocker and E. P. Simoncelli, Nature Neuroscience 9, 578 (2006)

    2006

  54. [62]

    D. C. Knill and W. Richards, Perception as Bayesian Inference (Cambridge University Press, 1996)

    1996

  55. [63]

    C. J. Edelson, FlEye Reader, https://github.com/ edelsonc/FlEye_Reader (2024)

    2024

  56. [64]

    R. R. de Ruyter van Steveninck, W. Bialek, M. Potters, R. H. Carlson, and G. D. Lewen, in Natural and artifi- cial parallel computation, Proc. of the 5th NEC Research Symposium, edited by D. L. Waltz (SIAM, 1996) pp. 21– 41

    1996

  57. [65]

    S. Roy, S. R. Sinha, and R. R de Ruyter van Steveninck, Journal of Neuroscience 35, 6481 (2015)

    2015

  58. [66]

    Wertheimer, Zeitschrift f¨ ur Psychologie61, 161–265 (1912)

    M. Wertheimer, Zeitschrift f¨ ur Psychologie61, 161–265 (1912)

    1912

  59. [67]

    Hassenstein and W

    B. Hassenstein and W. Reichardt, Zeitschrift f¨ ur Natur- forschung B 11B, 513 (1956)

    1956

  60. [68]

    J. C. Lagarias, J. A. Reeds, M. H. Wright, and P. E. Wright, SIAM Journal on Optimization 9, 112 (1998)

    1998

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.