Vision and Color Fall Data Blast: Session II

Tuesday, October 12, 12:30 – 14:30 ET // Register here

This event has already occurred. You can watch a recording here.

Presenters:

• In-vivo classification of human cone photoreceptors reveals crystalline S-cone sub-mosaics in the central retina presented by Sierra Schleufer, University of Washington

• Tissue properties of optic radiations representing the foveal and peripheral visual fields presented by John Kruper, University of Washington

• Motion/direction-sensitive thalamic neurons project extensively to the middle layers of primary visual cortex presented by Jun Zhuang, Allen Institute for Brain Science

• Two-dimensional shape perception is based on a salience map presented by George Sperling, University of California, Irvine

• Evidence for a dipper effect in the perception of global form: findings from psychophysics and EEG presented by Martin T.W. Scott, The University of York

• Visual Search in Virtual Reality (VSVR): A visual search toolbox for virtual reality presented by Jacob Hadnett-Hunter, University of Bath

• Perceptual brightness scales in a White's effect stimulus are not captured by multiscale spatial filtering models of brightness perception presented by Joris Vincent, Technische Universität Berlin

• Attentional modulation of early cortical chromatic responses presented by Mackenzie V. Wise, University of Nevada, Reno

• Effects of event number and adaptation duration on blur and face aftereffects presented by Idris Shareef, University of Nevada, Reno

• Three new models to account for impact of racially-heterogenous face-diet on face expertise presented by Ipek Oruc, University of British Columbia

• Probabilistic visual processing in humans and recurrent neural networks presented by Nuttida Rungratsameetaweemana, The Salk Institute for Biological Studies

• The mechanisms underlying enhanced auditory motion perception in early blind individuals presented by Woon Ju Park, University of Washington

• Sensorimotor synchronization with visual, auditory, and tactile modalities presented by Simon Whitton, University of Nevada, Reno

• BRENCH: An open-source framework for b(r)enchmarking brightness models presented by Lynn Schmittwilken, Universität Berlin

Moderators:

• Stacey Choi, Ohio State University

• Ravi Jonnal, University of California, Davis

Each data blast session feature a series of short talks followed by lots of time for questions and discussion. This data blast is being hosted by the Vision and Color Technical Division and the Fall Vision Meeting Planning Committee.

Abstracts:

In-vivo classification of human cone photoreceptors reveals crystalline S-cone sub-mosaics in the central retina

Presenter: Sierra Schleufer, Department of Ophthalmology, University of Washington, Seattle, WA, United States

Co-authors: Vimal Pandiyan, Department of Ophthalmology, University of Washington, Seattle, WA, United States. ; Palash Bharadwaj, Department of Ophthalmology, University of Washington, Seattle, WA, United States. ; Department of Ophthalmology, University of Washington, Seattle, WA, United States. , Department of Ophthalmology, University of Washington, Seattle, WA, United States. ; Ramkumar Sabesan, Department of Ophthalmology, University of Washington, Seattle, WA, United States

The topography of S-cones in the macula sets the neural constraints for coding the short-wavelength spectrum of color vision. We find that S-cones tile the central human retina with a non-random crystalline arrangement. This finding departs from previous studies, likely due to limited sampling. In 2 subjects we classified cones using Adaptive Optics Line-scan OCT and a bleaching stimulus of 660±10 nm. 8 ROIs per subject were classified at 1.5° and ~4° eccentricity across the 4 meridians. Numbers of total and S- cones per ROI spanned 541-3545 (mean: 1823) and 38-171 (mean: 99), respectively. We measured S-cone spacing in each ROI using the established method of Density Recovery Profile (DRP). To compare with random arrangement, we generated 1000 Monte Carlo (MC) simulations of each ROI such that its cone locations were maintained but locations of S-cones within it were randomized. We then measured the radius in units of inter-cone distance for which S-cone density was significantly lower than MC distributions, finding low density in a 1-cone radius in 13/16 ROIs (8/8 at 1.5°, p ≤ .037. 5/8 at 3.5°-4.5°, p ≤ .002), and up to a 2-cone radius in 12/16 ROIs (7/8 at 1.5°, p ≤ .002. 5/8 at 3.5°-4.5° p ≤ .003). Further experiments will include additional human subjects, ROIs at higher eccentricities, and classification using a short-wavelength bleach. Together, these findings have important implications for retinal development and color coding retinal circuits.

Funding acknowledgement: NIH U01EY025501, R21EY027941, R01EY029710, R01EY028118, P30EY001730 Research to Prevent Blindness, Research to Prevent Blindness Career Development Award.

Tissue properties of optic radiations representing the foveal and peripheral visual fields

Presenter: John Kruper, Department of Psychology, University of Washington, USA; eScience Institute, University of Washington, USA

Co-authors: Noah C. Benson, eScience Institute, University of Washington, USA; Sendy Caffarra, Graduate School of Education and Department of Pediatrics, Stanford, USA; Graduate School of Education and Department of Pediatrics, Stanford, USA, Graduate School of Education and Department of Pediatrics, Stanford, USA; Julia Owen, Department of Ophthalmology, University of Washington, USA; Yue Wu, Department of Ophthalmology, University of Washington, USA; Aaron Lee, Department of Ophthalmology, University of Washington, USA; Cecilia Lee, Department of Ophthalmology, University of Washington, USA; Jason Yeatman, Graduate School of Education and Department of Pediatrics, Stanford, USA; Ariel Rokem, Department of Psychology, University of Washington, USA; eScience Institute, University of Washington, USA

The biological properties of the foveal and peripheral visual pathways differ substantially. Here, we compared tissue properties of optic radiations (OR) carrying foveal and peripheral information to primary visual cortex (V1), measured with diffusion MRI (dMRI). We analyzed dMRI in two datasets: the Human Connectome Project (HCP; n=180; age 22-35) and the UK Biobank (UKB; n=7,088; age 45-81). In the HCP, OR was delineated using a three-dimensional atlas; parts of OR representing fovea and periphery were divided based on V1 fMRI responses close to OR endpoints. (2) In the UKB, OR was delineated using landmarks and divided using anatomically-based estimates of V1 responses close to OR endpoints. The dMRI signal was modeled using a kurtosis model, which provides information about tissue microstructure. Despite differences in data collection, population characteristics, and analysis methods, both datasets revealed higher fractional anisotropy, lower mean diffusivity, and higher mean kurtosis in the foveal OR than in peripheral OR, consistent with denser nerve fiber populations in foveal pathways. In further analysis of the UKB, we found that age is associated with increased diffusivity and decreased anisotropy and kurtosis, consistent with decreased density and tissue organization with aging. However, anisotropy in fovea decreases faster with age than in periphery, while diffusivity increases faster in periphery, suggesting foveal and peripheral OR differ in terms of aging.

Funding acknowledgement: Unrestricted and career development award from RPB, NEI/NIH K23EY029246 and NIA/NIH U19AG066567. AFQ projects supported through grant 1RF1MH121868-01 from the NIMH/The BRAIN Initiative.

Motion/direction-sensitive thalamic neurons project extensively to the middle layers of primary visual cortex

Presenter: Jun Zhuang, Allen Institute for Brain Science

Co-authors: Yun Wang, Allen Institute for Brain Science; Naveen D. Ouellette, Allen Institute for Brain Science; Allen Institute for Brain Science, Allen Institute for Brain Science; Emily Turschak, Allen Institute for Brain Science; Rylan S. Larsen, Allen Institute for Brain Science; Kevin T. Takasaki, Allen Institute for Brain Science; Tanya L. Daigle, Allen Institute for Brain Science; Bosiljka Tasic, Allen Institute for Brain Science; Jack Waters, Allen Institute for Brain Science; Hongkui Zeng, Allen Institute for Brain Science; R. Clay Reid, Allen Institute for Brain Science

The motion/direction-sensitive and location-sensitive neurons are two major functional types in mouse visual thalamus that project to the primary visual cortex (V1). It has been proposed that the motion/direction-sensitive neurons mainly target the superficial layers in V1, in contrast to the location-sensitive neurons which mainly target the middle layers. Here, by imaging calcium activities of motion/direction-sensitive and location-sensitive axons in V1, we find no evidence for these cell-type specific laminar biases at population level. Furthermore, using a novel approach to reconstruct single-axon structures with identified in vivo response types, we show that, at single-axon level, the motion/direction-sensitive axons have middle layer preferences and project more densely to the middle layers than the location-sensitive axons. Overall, our results demonstrate that Motion/direction-sensitive thalamic neurons project extensively to the middle layers of V1, challenging the current view of the thalamocortical organizations in the mouse visual system.

Funding acknowledgement: NINDS R01NS104949, NIMH R01MH117820

Two-dimensional shape perception is based on a salience map

Presenter: George Sperling, Department of Cognitive Sciences/University of California, Irvine, USA

Co-author: Lingyu Gan, Department of Cognitive Sciences/University of California, Irvine, USA

A salience map is a dynamic topographical map that combines information from individual feature maps into a real-number measure of conspicuity (salience). Originally, Koch and Ullman (1985) used salience to predict the priority of locations in visual search. The defining characteristic of a salience map is substance indifference--computations made on the map contents are independent of the features that produced the salience values because the features are not represented in the map. Salience maps have been proposed for computations other than search or processing priority: attention-based motion perception (Lu and Sperling, 1995), isoluminant red-green grating motion (Lu, Lesmes, Sperling centroid computation (Sun, Chubb, Wright, Sperling, 2018) and frontal-plane distance perception (Gan, Sun, Sperling, 2021). Here we demonstrate the well-known fact that simple shapes, such as letters and numbers, can be recognized not only when they are painted black or white but also when the are defined by outlines, or filled with different textures than the background, or filled with colors that are isoluminant with the background. Substance indifference means 2D shape perception is a computation performed in large part on a salience map. We further demonstrate that luminance is not necessary for accurate shape perception, that isoluminant text is easily readable, and so-called luminance artifacts are irrelevant, i.e., salience is sufficient for accurate 2D shape perception.

Evidence for a dipper effect in the perception of global form: findings from psychophysics and EEG

Presenter: Martin T.W. Scott, Department of Psychology, The university of York

Co-authors: Alex R. Wade, Department of Psychology, The university of York; Daniel H. Baker, Department of Psychology, The university of York; Department of Psychology, The university of York, Department of Psychology, The university of York; Heidi A. Baseler, Department of Psychology, The university of York

Decades of psychophysical experiments have shown that the perception of low luminance contrast violates Webers’ law: contrast discrimination is best at low (but non-zero) pedestal intensities. This “dipper effect” is thought to be the product of a sigmoidal neuronal transducer function in the early stages of visual processing which is expansive at low contrasts. Here, we ask 1) whether the transducer that governs the perception of global form is subject to a similar nonlinearity and 2) if sensitivity is biased towards a certain axis of pattern alignment (translational, radial, concentric). Using a combination of psychophysics and steady-state VEPs, we examined observers’ sensitivity to global form by manipulating the dipole orientation coherence of Glass patterns. Our psychophysical data indicate a Glass pattern “dipper effect” that is strongest for concentric patterns, while our EEG results show some evidence of nonlinear transduction but do not have the form predicted by psychophysical thresholds. Our findings indicate that, like low-level contrast, mid-level form discrimination is subject to mild facilitation (relative to detection) at low global form coherences, and suggest that this could be driven by neurons with nonlinear transducer functions.

Visual Search in Virtual Reality (VSVR): A visual search toolbox for virtual reality

Presenter: Jacob Hadnett-Hunter, Department of Computer Science, University of Bath, UK

Co-authors: Eamonn O'Neill, Department of Computer Science, University of Bath, UK; Michael J. Proulx, Department of Psychology, University of Bath, UK; Department of Psychology, University of Bath, UK, Department of Psychology, University of Bath, UK

Our understanding of human visual attention has greatly benefited from a wealth of visual search studies conducted over the past few decades. Task observers with searching for a specific target embedded among a set of distractors, and the time taken for them to find the target can reveal much about the sensory and cognitive processes involved. These experiments have typically been conducted on 2D displays under tightly controlled viewing conditions. Recently however, there have been calls within the visual attention community to explore more ecologically valid means of data collection. Virtual reality (VR) is a promising methodological tool for such research as it offers improved visual realism and the possibility of participant interaction, while retaining a significant amount of the control afforded by a computerized and monitor presented experiment. Here we present the Visual Search in Virtual Reality (VSVR) ToolBox. VSVR is a set of functions, scripts and assets that can be combined within a visual scripting environment in the Unity game engine to design, replicate and extend visual search experiments in VR. We further demonstrate the utility of such a toolbox with three experiments: a replication of feature search behavior, a demonstration of wide field-of-view visual search and eccentricity effects, and replication of depth plane as a feature for search.

Perceptual brightness scales in a White's effect stimulus are not captured by multiscale spatial filtering models of brightness perception

Presenter: Joris Vincent, Technische Universität Berlin, Germany

Co-authors: Guillermo Aguilar, Technische Universität Berlin, Germany; Marianne Maertens, Technische Universität Berlin, Germany; Technische Universität Berlin, Germany, Technische Universität Berlin, Germany

White's effect describes a brightness difference between two gray targets embedded in the black and white phases of a high contrast (square-wave) grating. Lacking an agreed explanation of this effect, it is often used as a target for computational models of brightness perception. Such models provide transfer functions linking input luminance to model output; if interpreting the latter as perceptual magnitude, these transfer functions are analogous to perceptual scales. However, whether the model transfer functions correspond to the perceptual scales has not yet been investigated. Here we estimate perceptual scales for White’s stimulus. On each trial, observers judged which of two targets appeared lighter. Each target varied in luminance across trials, and could appear either on the black or white phase of the grating. Using Maximum Likelihood Conjoint Measurement, perceptual scales were estimated from these judgments. Scales were compressive non-linear functions. Overall, the scale for targets on the black phase is shifted up relative to the white-phase targets (in agreement with the direction of White’s effect). This shift may be larger for intermediate luminance values than when target luminance approaches the luminance of the white phase, which has also been reported using more common matching tasks. Computational models failed to predict the shapes of the scales, suggesting that perceptual scales in addition to matching data could better constrain brightness models.

Funding acknowledgement: This work has been supported by research grants of the German Research Foundation DFG MA5127/3-1 and MA5127/4-1

Attentional modulation of early cortical chromatic responses

Presenter: Mackenzie V. Wise, Department of Psychology, University of Nevada, Reno, USA

Co-authors: Osman B. Kavcar, Department of Psychology, University of Nevada, Reno, USA; Alex J. Richardson, Department of Psychology, University of Nevada, Reno, USA; Department of Psychology, University of Nevada, Reno, USA, Department of Psychology, University of Nevada, Reno, USA; Michael A. Crognale, Department of Psychology, University of Nevada, Reno, USA

Prior research demonstrates that attention can modulate the amplitude of achromatic (black and white) pattern reversal stimuli. Similarly, MRI research has shown attentional modulation of BOLD responses to both chromatic and achromatic stimuli. However, prior research has also demonstrated that the chromatic onset visual evoked potential (VEP) is robust to attentional modulation (e.g., Highsmith & Crognale, 2010; Wang & Wade, 2011) with either spatially contiguous or spatially disparate manipulations. It is possible that more demanding attentional tasks than those that have been used previously, may reveal attentional modulation of the chromatic VEP. Here we report the results of experiments in which we recorded chromatic and achromatic VEPs while subjects performed multiple object tracking (MOT) tasks with several levels of difficulty. Steady-state onset responses were recorded at 3hz for stimuli modulated along the chromatic and achromatic axes. Our preliminary results suggest that the increased attentional demand provided by MOT can reveal a clear, but idiosyncratic modulation of chromatic VEP responses. The question remains whether or not chromatic and achromatic mechanisms share similar attentional gain processes as reflected in the VEP. We are currently characterizing the nature of these attentional gain mechanisms.

Effects of event number and adaptation duration on blur and face aftereffects

Presenter: Idris Shareef, Department of Psychology, University of Nevada, Reno, USA

Co-authors: Mohana Kuppuswamy Parthasarathy, Department of Psychology, University of Nevada, Reno, USA; Michael A Webster, Department of Psychology, University of Nevada, Reno, USA; Department of Psychology, University of Nevada, Reno, USA, Department of Psychology, University of Nevada, Reno, USA; Alireza Tavakkoli, Department of Computer Science and Engineering, University of Nevada, Reno, USA; Fang Jiang, Department of Psychology, University of Nevada, Reno, USA

Previous studies have revealed that the strength of adaptation aftereffects can depend on both the number of adaptation events (trials) and the duration of each event. However, how these effects depend on the stimulus property adapted remains unknown. In the present study, we compared the influence of adaptation event number vs event duration on the strength of adaptation for blur or face aftereffects. For blur adaptation, we filtered a natural image’s amplitude spectrum over slopes between -1 (blurred) to +1 (sharpened) relative to the original image to create a series of blurred/sharpened images. For face adaptation, we morphed an Asian and Caucasian face image resulting in a finely graded series of images spanning the two ethnicities. During each top-up adapting period, we varied the number of adaptation events (4 or 16) and duration of each event (250ms or 1s), resulting in 4 event number-event duration conditions. Our results suggest that the effects of event number and event duration on the strength of aftereffects are similar for blur and face adaptation.

Funding acknowledgement: P20 GM103650, FA9550-21-1-0207

Three new models to account for impact of racially-heterogenous face-diet on face expertise

Presenter: Ipek Oruc, Department of Ophthalmology and Visual Sciences, University of British Columbia, Canada

Co-author: Morteza Mousavi, Department of Ophthalmology and Visual Sciences, University of British Columbia, Canada

Racially-homogeneous environments provide limited experience with other-race faces (Oruc et al. 2019). According to the contact hypothesis, this, in part, leads to an impairment in other-race face recognition yet does not predict what impact a racially-heterogenous face-diet may have on face expertise. To complement and extend the contact hypothesis, we propose three new models: (1) the experience-limited, (2) the capacity-limited, and (3) the enhancement hypotheses, for the role of exposure in face expertise. Based on the experience-limited account native-level face recognition can be achieved for multiple races with sufficient experience. The capacity-limited account predicts exposure to multiple races may impact face expertise detrimentally. Lastly, the enhancement account posits advantages of a racially-heterogenous face-diet. In two experiments, we compared face recognition in a dual-exposure group with sustained high exposure to Caucasian and East Asian faces to two respective mono-exposure groups. We found native-like recognition in the dual-exposure group for both Caucasian and East Asian faces. Our results showed neither an advantage, nor a disadvantage for racially-heterogenous face exposure, hence supporting the experience-limited account of face expertise. We conclude that exposure to multiple face races is not detrimental to face recognition ability. To achieve native-level face expertise, a racially-homogenous face diet is not required.

Funding acknowledgement: NSERC Discovery Grant RGPIN-2019-05554; an Accelerator Supplement RGPAS-2019-00026, and a Canada Foundation for Innovation, John R. Evans Leaders Fund.

Probabilistic visual processing in humans and recurrent neural networks

Presenter: Nuttida Rungratsameetaweemana, Computational Neurobiology Laboratory, The Salk Institute for Biological Studies, USA

Co-authors: Robert Kim, Computational Neurobiology Laboratory, The Salk Institute for Biological Studies, USA; John T. Serences, Department of Psychology, University of California San Diego, USA; Department of Psychology, University of California San Diego, USA, Department of Psychology, University of California San Diego, USA; Terrence J. Sejnowski, Department of Biological Sciences, University of California San Diego, USA

Visual inputs are often highly structured, and statistical regularities of these signals can be used to guide future visuomotor associations and thus optimize behavior. Through a recurrent neural network (RNN) model, human psychophysics, and electroencephalography (EEG), we probed the neural mechanisms for processing probabilistic structures of visual signals to guide behavior. We first constructed and trained a biophysically constrained RNN model to perform a series of probabilistic visual discrimination tasks similar to paradigms designed for humans. Specifically, the training environment was probabilistic such that one stimulus was more probable than the others. We showed that both humans and RNNs successfully learned the stimulus probability and integrated this knowledge into their decisions and task strategy in a new environment. Performance of both humans and RNNs varied with the degree to which the stimulus probability of the new environment matched the formed expectation. In both cases, this expectation effect was more prominent when the strength of sensory evidence was low, suggesting that like humans, the RNNs placed more emphasis on prior expectation (top-down signals) when the available sensory information (bottom-up signals) was limited, thereby optimizing task performance. By dissecting the trained RNNs, we demonstrated how competitive inhibition and recurrent excitation form the basis for neural circuitry optimized to perform probabilistic visual processing.

Funding acknowledgement: NIMH (F30MH115605-01A1 to R.K.); NEI R01EY025872-10 (J.T.S.); NIBIB R01EB026899-01 (T.J.S.); NINDS R01NS104368 (T.J.S.); and Mission funding from the U.S. Army Research Laboratory (N.R.)

The mechanisms underlying enhanced auditory motion perception in early blind individuals

Presenter: Woon Ju Park, Department of Psychology, University of Washington, USA

Co-author: Ione Fine, Department of Psychology, University of Washington, USA

Introduction. Previous studies have shown increased sensitivity to auditory motion in early blind (EB) individuals. However, the mechanisms underlying this enhanced performance remains unknown. We estimated auditory motion filters in EB using a psychophysical reverse correlation paradigm—the auditory analogue of Neri (2014).

Methods. Participants (8 EB and 8 sighted controls (SC)) discriminated the direction of a signal motion (left/right) embedded in broadband noise bursts (500-14,000 Hz) presented over a 10x10 grid (space: -/+30 °; time: 0-800 ms). Amplitude of signal motion was adjusted using staircases to maintain performance at 65%. We made two independent measurements: tuning of auditory motion filters (how noise bursts influence the probability of hearing motion direction), and amplitude threshold for hearing the signal.

Results. The estimated filters for both EB and SC were predominantly responsive to sound onsets/offsets, different from the non-separable spatiotemporal selectivity previously observed for visual motion. Also, the filters for EB were more accurate in detecting signal onsets/offsets across both space and time. This more refined tuning successfully predicted the greater sensitivity to hear signal motion in EB, as measured by the amplitude threshold (t(14) = -5.59, p < .001). Thus, EB individuals show subtle qualitative rather than quantitative differences in auditory motion processing that nonetheless result in significant improvements in perception.

Funding acknowledgement: Weill Neurohub Postdoctoral Fellowship to WP; NEI R01 EY014645 to IF.

Sensorimotor synchronization with visual, auditory, and tactile modalities

Presenter: Simon Whitton, Department of Psychology, University of Nevada, Reno, USA

Co-author: Fang Jiang, Department of Psychology, University of Nevada, Reno, USA

While it is well known that humans are highly responsive to musical rhythm, the factors that influence our innate ability to synchronize remain unclear. In the current study, we examined how stimulus complexity and modality, along with the synchronizer’s level of musicality, impacted sustained sensorimotor synchronization (SMS). Utilizing a finger-tapping task and three sensory modalities (visual, auditory, and tactile), we manipulated rhythmic complexity by varying the location and availability of temporal cues across four conditions. Additionally, to determine our participants’ (n = 30) musical experience and aptitude, we administered the Goldsmiths Musical Sophistication Index (Gold-MSI) questionnaire. We found that SMS to external rhythmic stimuli was significantly more precise for auditory and tactile than for visual sequences. Further, we found SMS precision significantly decreased in all modalities as rhythmic complexity increased, suggesting rhythmic complexity directly relates to SMS difficulty. Moreover, a significant correlation was found between Gold-MSI scores and SMS accuracy in the most rhythmically complex condition, such that the higher one’s musicality score, the greater one’s accuracy. This held for all three modalities. Combined, these findings suggest that rhythmic synchronization performance is affected not only by the modality and complexity of the rhythmic stimuli but also by the musicality of the synchronizer.

Funding acknowledgement: P20 GM103650

BRENCH: An open-source framework for b(r)enchmarking brightness models

Presenter: Lynn Schmittwilken, Exzellenzcluster Science of Intelligence, Technische Universität Berlin, Germany

Co-authors: Matko Matic, Exzellenzcluster Science of Intelligence, Technische Universität Berlin, Germany; Marianne Maertens, Exzellenzcluster Science of Intelligence, Technische Universität Berlin, Germany; Exzellenzcluster Science of Intelligence, Technische Universität Berlin, Germany, Exzellenzcluster Science of Intelligence, Technische Universität Berlin, Germany; Joris Vincent, Department of Computational Psychology, Technische Universität Berlin, Germany

Various image-computable models have been proposed to describe the relationship between local luminance, visual context, and perceived luminance (brightness) or perceived reflectance (lightness). Classically, these models are tested on a subset of relevant stimuli. While specific failure cases have been shown for most models, a systematic overview is lacking. As a consequence, it is unclear how to favor any specific model of human brightness perception.

Our goal is to work towards such a comprehensive overview. Towards that end, we are developing a stimulus benchmark, and evaluate various brightness models on these stimuli. For this, we provide publicly available re-implementations of several brightness models in Python, code for creating parameterized stimuli, and BRENCH - a framework to automate running and evaluating brightness models on (sets of) stimuli. With our framework, we can replicate previously published modeling results. Going beyond, the framework facilitates the comparison of models from multiple publications across all stimuli from those publications. BRENCH is flexible to allow for new model(parameterization)s and stimuli. Comparing a larger set of models and stimuli makes it possible to group models which perform similarly on certain stimuli, and group stimuli based on similar model performances. We hope BRENCH aids discussions about what stimuli should form a benchmark for brightness models, and what the interface and form of such models should be.

Funding acknowledgement: Funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy – EXC 2002/1 “Science of Intelligence” – project number 390523135.