01 Dec Student Awards 2021
Wednesday 1st December 2021 14.00-16.00
Please sign up at eventbrite link below.
Every year the Colour Group (GB) supports a number of outstanding student and Early Career postgraduate members with financial support in the form of awards. These awards allow early career researchers to attend international conferences in order to present their work, and also recognises outstanding posters at a number of conferences. Awardees are invited to share their work with members and this year we are delighted to be able to host the award holders’ talks for members in this online event.
The award winners will be presenting their award-winning research in an online seminar. Attendance is open to members of the colour group as well as non-members. Abstracts are provided below.
Please sign up to attend the online event at the eventbrite link below. Abstracts are below the sign-up link
14.00 Introduction
14.05 Gyeonghwa Lee – University of Leeds
The role of colour designers in the design process – WD Wright Award
14.30 Allie Hexley – University of Oxford
A machine-learning approach to illuminant estimation using statistical regularities in photoreceptor signals from real-world surfaces – WD Wright Award
14.55 Break
15.05 Manuel Spitschan – University of Oxford
Retinal inputs to human neuroendocrine and circadian physiology: Role of S cones and rods – Palmer Award
15.30 Takuma Morimoto – University of Oxford
Modeling perceptual discrimination of surface color using image chromatic statistics and convolutional neural networks – Palmer Award
15.55 Closing
For further information concerning this year’s and next year’s awards, please contact the Colour Group Awards Officer: t.l.v.cheung@leeds.ac.uk
Abstracts
The role of colour designers in the design process
Gyeonghwa Lee, University of Leeds
[presented at the International Colour Association (AIC) 14th Congress]
The purpose of this study is to offer an understanding and knowledge about the role of colour designers in the design process (particularly paint industry), and to discuss the required attributes and skills of colour designers in each stage of the design process. In order to achieve the goals of the research, a qualitative autoethnography was conducted based on five years of experience gained by one of the authors as a colour designer in paint companies. The experience was described in line with five-stages for a colour design framework based on the idea of general tendencies in the problem-solving process. This research provides a deeper understanding and knowledge about the work of colour designers in the practical design environment, as well as providing a clear colour design framework based on the practical activities of colour designers. The research is expected also to increase awareness of the roles of colour designers in our society. Increased awareness may lead designers to enter the colour design area and engage in colour design research.
A machine-learning approach to illuminant estimation using statistical regularities in photoreceptor signals from real-world surfaces
Allie Hexley, University of Oxford
[presented at the 43rd European Conference on Visual Perception]
In natural scenes, illuminant colour can provide biologically relevant information about time-of-day, and local conditions of shade and heading. Illuminant estimation is also a vital step in many models of colour constancy, in which surfaces maintain stable colour appearance under changing illumination. Yet, in a world without physical constraints, a single sample of reflected light from a surface could in principle be produced from an infinite set of possible surface-illuminant pairs, and illuminant estimation is mathematically under constrained. Here we use a machine-learning approach to quantify, for a dataset of real-world surface and illuminant spectra, the limits that can be placed on illuminant estimation from photoreceptor signals. We reduced the radiance spectra to five photoreceptor coordinates (the three cone classes, the rods, and melanopsin), which are in principle available to a human observer. Algorithms were trained on the combination of photoreceptor signals associated with 49,667 real-world surfaces labelled by the illuminant under which the surfaces were placed. We evaluate the properties of the surfaces, illuminants, and photoreceptor signals that impact performance of the algorithms to predict illuminant chromaticity. For example, when trained using real-world surfaces under five daylight illuminants of different colour temperature, and using a combination of all five photoreceptor signals, we find our classifiers have high accuracy, precision, and recall at correctly identifying the illuminant from an unseen set of photoreceptor signals from a single surface. The approach also allows us to quantify the advantage gained from combining signals from multiple surfaces, and the optimal strategy of combination.
Retinal inputs to human neuroendocrine and circadian physiology: Role of S cones and rods
Manuel Spitschan, University of Oxford
[presented at the 85th Cold Spring Harbor Laboratory Symposium on Quantitative Biology: Biological Time Keeping]
In addition to canonical visual functions, light exposure profoundly affects physiology and behaviour through its effects on circadian, neuroendocrine and neurocognitive functions. This is largely mediated by the melanopsin-containing intrinsically photosensitive retinal ganglion cells (ipRGCs). However, ipRGCs also receive cone-rod input, making the signal driving the non-visual effects of light complex. Here, I will discuss two lines of my work investigating how the different photoreceptors in the retina impact human neuroendocrine and circadian physiology. First, I will describe a recent study examining the role of S cones in melatonin suppression by evening light, employing the method of silent substitution to generate spectral light stimuli that differ by almost two orders of magnitude in S cone activation with minimal change in the activation of the L and M cones, melanopsin and rods. Second, I will describe a recent field study with congenital achromats, who do not have a functional cone system, examining the regularity of their rest-activity cycles using actimetry, and the melatonin phase angles of circadian entrainment.
Modeling perceptual discrimination of surface color using image chromatic statistics and convolutional neural networks
Takuma Morimoto, University of Oxford
[presented at the Vision Science Society Meeting 2021]
A previous study measured thresholds to discriminate colors of objects under each of three different lighting environments. Discrimination thresholds were similar for matte and glossy objects, but the orientations of the discrimination ellipses were tightly aligned with the chromatic variation of the lighting environment in which the objects were placed (Morimoto & Smithson, 2018). Using two distinct modeling approaches we analyzed the psychophysical data to reveal the potential strategies that humans used to perform the discrimination task. First, we built three hand-crafted models that discriminate objects’ colors by comparing specified chromatic statistics: (i) mean chromaticity, (ii) chromaticity of the brightest pixel, and (iii) luminance-weighted-mean chromaticity. In the second approach, we trained convolutional neural networks (CNNs), based on 38,021 images labelled either by physical ground-truth or human responses. Then, thresholds were estimated for all models using the identical staircase procedure that measured human thresholds. The first approach showed that the mean chromaticity and the luminance-weighted-mean-chromaticity models predicted human thresholds generally well, but the brightest-pixel model predicted thresholds better in some matte conditions, indicating that no tested model based on single chromatic statistics can predict thresholds consistently well across conditions. Moreover, the estimated thresholds for these models were generally higher than human thresholds. In contrast, the CNN trained on images with human responses nearly perfectly predicted human thresholds in all conditions. The CNN trained on physical ground-truth showed much lower thresholds than human thresholds. Visualizing activation maps of the CNN trained on human responses revealed that the CNN primarily looks at shaded regions of the surface that are dominated by diffuse reflections of indirect illumination and thereby provide more reliable information about surface color. Combining hypothesis-based and data-driven approaches revealed an effective strategy to separate lighting and material to reliably discriminate surface color under complex lighting environments, which humans might also use.