Over the last ten years, X-Rite's Appearance Exchange Format (AxF) has been increasingly adopted across various industries for sharing digitized material appearances among departments, suppliers, and customers. Although not an oficial standard, AxF's neutrality (with regard to rendering applications) and the absence of a suitable industry standard have led many rendering applications to include support for the format. In particular, the AxF ecosystem supports the creation of truly trustworthy digital twins of real world materials by combining a well-defined and documented set of physically based material shading models with validation procedures to verify the shading models against calibrated measurement data. It also connects to existing and well-proven workflows based on established standards like color appearance models or gloss units.
B2B
B2C
Material Supplier Manufacturer/Brand
Point of Sale
The pressures of a globalized economy are manifold: increased competition, shorter product life cycles,
supply chain disruptions, and regulatory compliance requirements, to name a few. Equally numerous
are the promises of digitalization: waste and cost reduction, increased eficiency, faster response times,
quality improvements, etc. Consequently, more and more industries are introducing digital twins into
their product development and production processes. The concept of the digital twin was first formally
introduced by Michael Grieves (e.g., [
The complexity and diversity of color and material appearances poses significant challenges in achieving a consistent and accurate digital representation of material appearance. This is especially true for industries such as automotive, consumer goods, fashion and architecture, where the visual quality and consistency of material appearance plays a crucial role in product design and customer satisfaction. Additionally, suppliers are located in diferent regions of the world, and the materials used often exhibit fundamentally diferent appearance characteristics.
Consequently, there currently exists a plethora of diferent approaches and tools tailored to the
specific demands of each industry. We argue that many of these existing approaches fall short of
facilitating the eficient communication of material appearance and can lead to misunderstandings and
misinterpretations, which might result in costly mistakes and delays. Typical examples include the
following:
• Physical sample swatches - are still the most common way to communicate material appearance,
but they need to be produced, shipped and stored, which is time consuming and costly. Trying
to imagine from a small swatch how the material will appear on the actual product is another
challenge that can lead to unexpected outcomes. Furthermore, the samples are subject to aging
and degradation, which can result in inconsistencies over time.
• "Magic" numbers - like SKUs, RGB values or gloss levels. This approach is widely used in industries
where a huge number of material variations, like colorways for example, is ofered and producing
physical sample swatches for all variations is simply not feasible. However, single numbers are
not suficient to communicate the full appearance of a material. In practice, a lot of expertise
is required to interpret the numbers correctly, and comparing "number systems" from diferent
suppliers or manufacturers is often not possible.
• Metaphorical language - like "an earthy brown in the spirit of dried clay" is mainly used in design
departments, but is highly subjective and can lead to misunderstandings and misinterpretations.
• Uncalibrated images or videos - can easily be misinterpreted and in general do not provide a true
representation of the material’s color and appearance.
• Proprietary and artistic PBR material shaders - most of the material models used in 3D rendering
and visualization software packages are still crafted using uncalibrated methods like flatbed
scanning and artistic ("by eye") tuning of appearance parameters like gloss, transparency or
anisotropy. Despite recent initatives like OpenPBR [
We argue that many of the aforementioned problems can be mitigated by a verifyable and vendorneutral visual digital twin. It needs to be visual to facilitate intuitive communication - number codes only understood by experts are not acceptable. Verification is necessary to build trust. Otherwise, decisions wouldn’t be reliable and stakeholders would fallback to physical samples immediately. Last but not least, vendor-neutrality based on standardized data formats and protocols, is required to ensure interoperability and ease of use across diferent platforms and applications.
While to our knowledge there is no global standard for digital twins of material appearance, we would argue in the following that the Appearance Exchange Format (AxF) is a promising candidate, which fulfills many of the requirements for a visual digital twin. As a result, we see an increasing adoption of AxF in the industry, which could be understood as soft standardization.
One strength of AxF is that it builds on decades of experience gained in the management, measurement and communication of color, which is one of the most important visual properties of customer-facing materials.
In certain industries, the appearance of the final product is dominated and identified by color. An
example is the printing industry, where the color of the printed material is usually its most important
appearance aspect. Over the years, the industry developed a strong set of standards, devices, tools and
protocols to facilitate the digital handling of colors and specify an almost complete visual digital twin
of color:
• Color naming systems like RAL or Pantone provide a standardized way to identify and communicate
colors and while originally based on physical samples, they are now also available digitally [
Based on these components, as illustrated in Figure 2, a comprehensive digital representation of color can be achieved, enabling seamless communication and collaboration across diferent stages of the product lifecycle.
We sketched in the previous section that in some areas, like printing for example, a single dimension of appearance (like color) can be managed independently of others. Gloss, which can be measured by gloss meters and represented by perceptual gloss units, is another example where a set of standards and best practices does exist. However, only very few materials can faithfully be represented by a single appearance dimension like color or gloss.
In general, visual appearance emerges from the complex interaction between light and matter, and
its diferent dimensions are highly interdependent. In fact, looking at gloss and color alone reveals
that the perception of each dimension is highly dependent on the other (e.g. [
While the lack of holistic and practical representations for appearance indicates that there is still a lot
of fundamental research needed, the industry is pressured today and the need for the digitalization of
the appearance supply chain is real. We therefore propose the Appearance Exchange Format (AxF) [
In order to address the complexity of digital material appearance communication, the AxF formant has been designed to represent diferent aspects of material appearance at diferent levels of detail as illustrated in Figure 3.
K+S a t a d a t e M
Recipes Appearance Traits, Keys, “Factors”, … Appearance Rendering
Model Parameters Calibrated Measurements
Dense/Sparse/Spot •Production Parameters •Formulation (Pigment concentration, Assortment, …) •Albedo (diffuse reflection factor), Gloss Unit, Roughness Factor, TextureScales, K+S, etc. •Computed from Measurements/Appearance Models via Data Analysis •E.g., use for Quality Control •Optimized for Visualization •May contain data loosely connected to semantic (Roughness, Diffuse, …) •Overlap with rendering applications •Dense: TAC7 measurements •Sparse: MA-T12 •Spot: Ci7, MetaVue etc.
Calib Data
Raw Measurement Device Readings e c n a v e l e R s s e n i s u B / n o it c a r t s b A Legend: already implemented not implemented abc software component interlevel transformation
Formula on
Predic on
Data Analysis Fi ng
Rendering
Decoding
Calibra on
Normaliza on
The foundation of the AxF data pyramid represents accurate appearance measurement data. Storing measurement data in AxF is essential to enable the verification of the visual digital twin for trust building.
Please note, that the lowest level is not yet implemented and used in practice. As of now, the calibration and normalization of raw sensor readings is usually done on the measurement device and only the resulting calibrated measurement data is stored in AxF. However, since data storage and transmission costs are continuously decreasing, it is expected that in the future raw measurement and calibration data will also be stored in AxF to allow full traceability.
In general the pyramid’s foundation can be understood as an extended version of CxF, which cannot only represent colors and spectra but also images and other sensor readings like surface (height) profiles.
Spatially Varying Reflectance Distribution Functions (SVBRDF) Dielectric & Metallic Highlights Transparency Anisotropy Colored Transmission Gloss Colored Clear Coat Normal map Orange Peel Height map Sheen Car paint Color Table Flake BTF Clear Coat Orange Peel
Volumetric Scattering Absorption IOR Normal map
As mentioned in Section 3 the interaction of light and matter is a multi-dimensional process which
can generally be modeled by a radiative transfer equation, with the classical rendering equation [
(x, , )(x, )( · n)
Ω
where is the outgoing radiance at point x in direction , is the emitted radiance, is the
incoming radiance from direction , n is the surface normal, and is the spatially varying bidirectional
reflectance distribution function (SVBRDF) [
Figure 4 summarizes the material appearance models specified in AxF. The majority are so-called
(SV-)BRDF models like [
Last but not least, AxF also provides models for homogenous volumetric materials like plastic that can be described by (spectral) scattering and absorption coeficients. (2)
In order to facilitate the communication of material appearance on a higher level of abstraction, AxF also provides means to store appearance traits which are high-level abstractions of the material’s appearance, such as a dominant color, gloss units, or profile roughness parameters. These traits allow to connect to workflows which still communicate based on single colors or gloss units. These traits can be derived from the appearance measurements and the shading model parameters using data analysis, but they can also be specified manually by a material expert or determined with devices like gloss meters or specialized laboratory equipment.
The tip of the AxF pyramid is reserved for recipes, which are high-level descriptions of how to actually produce a material with a specific appearance. Recipes can include information about the material’s composition, manufacturing process, and other relevant details. While not yet fully implemented, the connection with, e.g., paint formulation software is natural and enables the creation of a complete digital twin of the material, including its production process.
Last but not least, AxF allows to store customizable metadata along all levels of detail.
Deriving the parameters of the (SV-)BRDF models introduced in Section 4.2 from appearance measurements is a complex machine learning task, known as data fitting. The inverse process, which evaluates the model for a given set of parameters, is called rendering and allows to create a visual digital twin of the material’s appearance. Given an accurate model of the original measurement device, this process can also be used for verification, i.e., to verify that the rendered appearance matches the original measurement data. The diference between the simulated and the measured appearance shall be quantified using perceptual appearance metrics that extend the well-known colour metrics introduced in Section 2. In many practical applications, though, this verification is still done by visual inspection, e.g., by comparing the rendered appearance with the original sample in a calibrated viewing booth as shown in Figure 5 (left).
Rendering (real-time)
Real Object
It is long known that tri-chromatic rendering is not suficient to accurately simulate colors lit by
spectrally diferent light sources (e.g. [
We have laid out that the Appearance Exchange Format (AxF) provides a solid foundation for the digital representation of material appearance, which can be used to create visual digital twins of materials. While AxF is not an oficial standard yet, its increasing adoption in diverse industries and at diferent parts of the supply chain (e.g., raw material suppliers, material suppliers, brands etc.) can be seen as a form of soft standardization. The adoption is mainly driven by the real need for the supply chain optimizations potential ofered by visual digital twins. By combining accurate appearance measurements, physically based rendering models, and customizable metadata, AxF enables the creation of trustworthy digital twins that can be used across diferent applications and platforms.
From our experience one of the main missing ingredients required for the future progress of visual digital twins and their integration into the supply chain is a perceptual appearance metric which generalizes color metrics like ∆ 76 (cf. Equation 1) to holistic appearance. Such a metric would allow to quantify the diference between an appearance standard and its rendered and physical versions for more eficient product development and manufacturing.
During the preparation of this work, the author(s) used Github Copilot in order to: Drafting content and Grammar and spelling check. After using these tool(s)/service(s), the author(s) reviewed and edited the content as needed and take(s) full responsibility for the publication’s content.