Leading the Innovation in Surface Digitization
15 minutes read
Meet the colormass Material Scanner that is unmatched in realism and throughput.
This article is part of a series written for individuals who would like to gain insight into the technical challenges of leaving behind the traditional way of marketing material creation by switching to the 3D approach. Computer graphics is a complex topic and the goal of this series is to give enough technical background to allow for appropriate business decisions, while also not overwhelming the reader with unnecessary details or mathematical equations.
When working on high quality visuals in 3D, the quality of the underlying source data is of paramount importance. The creation and management of this data poses a significant challenge to the industry. As described in our previous article on the Basic Building Blocks of 3D Objects, there are two main types of information needed to create visuals: models and materials. Textiles fall within the materials category.
From a computer graphics and technical point of view, textiles are one of the most complicated surfaces to capture and to represent in 3D. The interaction between light and a piece of fabric is astoundingly complex. The intricacy of material capture stems from the properties of a textile – not only special weaving constructions but also the characteristics of the yarns from which they are made. Cotton, nylon, silk, wool or metallic threads all have different properties and varying “hair”-like microstructures.
To accurately represent materials digitally, the scanning method must be sophisticated enough to capture all of the textile’s complex characteristics – color, weave, fibers, sheen or similar. The method must capture all data needed to reproduce the intricate details of a real-world surface from any angle in any lighting condition on any piece of furniture. This level of complexity is the main reason digitizing a textile using either a flat-bed scanner or by simply photographing it from above on a flat surface will not provide enough detail to produce visually enticing results in 3D. These methods capture a flat sample from a single angle in a single lighting condition.
The above images of Bloomer by Designtex show the difference between the two material capture methods in 3D. On the left, the visual is produced using the flat tileable asset created from a top-down photograph. The image on the right uses materials digitized by colormass's sophisticated material capture methods. The difference in visual detail and quality is why the textile industry is quickly adapting this advanced scanning approach.
Keep in mind that this next generation approach of surface digitization is only possible due to recent technological advancements in the field of computer graphics. Historically, the entire textile industry relied on flat tiles for catalog, website and other marketing purposes. A flat tile is a regular digital image created by photographing a textile sample laid flat on a table. For reference, you can find an example of flat tiles from Kvadrat’s website below.
Besides providing imagery for collections, these tiles are also used by interior designers, architects or space planners in tools such as Revit, CET Designer, or pCon Planner. These flat photographs are useful, as they are tileable, meaning that the image can be placed adjacent to itself without noticeable seams. In design tools, this seamless aspect is necessary when applying them to furniture pieces to avoid having tiling artifacts, that is, visible seams between the tiles.
In the images above, the same furniture model has two 3D materials to showcase the difference between a non-tileable and tileable texture. Notice the repeating pattern distracting the viewer from the natural elegance of the product. While this is a good initial step, the tileable scans are not sufficient for high-quality visuals (images, configurators or videos) given the amount of information disregarded in the source data. Hence why the so-called PBR scans come into play.
PBR scans are considered to be the next generation of flat tiles. Not only do they tile seamlessly but also open up a whole new world of possibilities. PBR scans contain much more data on real-world, physical information regarding the actual look and structure of a textile than the standard tile. This is actually where the PBR name comes from: physically based rendering.
PBR does not adhere to a strict set of rules but it is rather a generic concept of physical correctness, so when simulating the real world in 3D, it looks to the laws of physics. As the well-known book Physically Based Rendering: From Theory To Implementation by Matt Pharr, Wenzel Jakob, and Greg Humphreys describes: “Physically based techniques attempt to simulate reality; that is, they use principles of physics to model the interaction of light and matter. While a physically based approach may seem to be the most obvious way to approach rendering, it has only been widely adopted in practice over the past 10 or so years.” Physically based rendering methods have produced a whole new level of visual quality and realism previously unachieved. If you are interested in learning more about the history of computer graphics, make sure to watch The Physics of Light and Rendering by John Carmack.
To be able to perform physically based rendering, one must capture data based on the physics of the material. Colormass realized 99% of the textile industry lacked the digital data required to produce physically based rendered images of a given caliber. Creating a proprietary system was essential and the team developed an in-house material capturing solution: a system of dozens of lights and cameras that work in concert with one another to capture the physical nature of the surfaces and colors when interacting with light. This scanning method, not to be confused with a flat-bed scanner, captures physical properties of a material from different angles and in different lighting conditions, superseding the more common tile scanning practices.
In addition to superior material data capturing, the colormass solution solves two issues in the digitization pipeline: sample capture size and volume of samples to be scanned. Most material capture solutions available today limit the size of physical samples, restricting the size of the pattern able to be captured. The colormass Scanner can handle large scale patterns, far exceeding the size of many of the alternative scanning solutions. Second, textile manufacturers typically offer hundreds of collections with a wide array of colorways and capturing physically accurate data for an entire library has been a large hurdle for many. The colormass scanning solution combats these pain points with their streamlined system.
The following section offers insight into the inner workings of our digitization process, detailing the steps of creating PBR scans and explaining the output and applications for businesses.
In order to scan a fabric, a physical sample of the material is required, which has to be shipped to our office. The sample should be around 0,5 by 0,5 meter/yard, but large enough to include the pattern repeat one and a half times.
Sample preparation and region selection. Once received, the initial step is to inspect how the shipping was handled, unpack the sample and place it carefully on the scanning area. Every fabric is unrolled top to bottom, the selvedge following the left and right edges of the work area. Although the pattern can be rotated later on in the rendering process when the material is applied to a piece of furniture, this ensures that the default orientation is always consistent. Also, this makes it easy to tell if a fabric is railroaded or not just by looking at the scan.
Due to the amount of data collected from the digitization process, every bump, crease, fold or imperfection will be visible. Pristine material samples are essential. Handled with care, white cotton gloves are worn as even fingerprints are detectable on certain materials like vinyls, metals, or plastics. After a quick overview scan, the region of interest is selected and the scan begins.
Image acquisition and pre-processing. The scan has started and the concert of lights commences. In a controlled, pitch-dark environment, the custom-designed light and camera system records more than 1,000 photos of the sample from different angles in different illuminations to gather the necessary data regarding how the surface interacts with light. The scanner captures these interactions in the highest detail technologically possible, taking advantage of HDR (High Dynamic Range) imaging and the newest, high-resolution industrial cameras and sensors available.
In order to achieve the most accurate results possible, the entire capture system has to be calibrated. The calibration consists of multiple steps. First, the exact positions of the lights and the cameras have to be determined in relation to each other and to the sample. Without this, it would be impossible to do any calculations. Second, the imperfections of the optical system have to be accounted for. The lens and homography calibrations not only eliminate barrel distortion but also deliver a pixel per millimeter coefficient, which allows for calculating the exact physical size of the digitized sample at submillimeter precision. This information is essential to match the 3D materials to the correct real-world size when applied to furniture pieces. Last but not least, the camera sensors are color calibrated, guaranteeing the highest color accuracy achievable with the most up to date hardware on the market.
Once the measurements are gathered, the data is pre-processed, compressed and uploaded into the cloud-based colormass Platform along with the corresponding calibration dataset.
Fitting. The fitting is the core of the digitization process. Our algorithms extract the physical properties of the surface, interpreting how the material reacts to light and evaluating the image data pixel by pixel. With fully proprietary technology developed by colormass, this computationally intensive task takes place in the cloud on a computer cluster using thousands of CPU cores and hundreds of gigabytes of RAM. For reference, if the same calculations were run on a regular laptop or desktop, it would take multiple days for a single fabric to be fitted. When calculated on the cluster, the process takes less than an hour to complete. This incredible speed is made possible by the heavy parallelization and clever task management of the processing pipeline. Once the fitting is completed, the data is passed over to the next stage of the pipeline in the form of texture maps.
As seen in the maps shown above, each of them contains data on the surface properties. For example, the diffuse map stores the actual color of the textile while the normal map describes the structure of the fabric, and the roughness map influences how glossy the surface appears. It is important to note that without a lot of experience in PBR shading, these maps will not mean much to the untrained eye. Nevertheless, it is helpful to think about them as the different visual attributes of a surface segmented into layers.
Tiling. PBR maps, like the flat images previously mentioned, must be tiled as well. Creating tileable maps is not an easy task, especially given that a PBR scan consists of several stacking layers which must align and work in unison. Furthermore, the human brain excels at recognizing patterns, especially in cases of solid fabrics (e.g. single colored woven linen) without a designed, repeating pattern like flowers, geometric shapes or similar jacquards. The smallest artifact—a fold in the fabric, a misaligned yarn or a small extruding lint—makes the tiling become immediately apparent, creating a false “pattern” disrupting the notion of aesthetics and realism. One possible method, albeit time consuming, to solve this problem is to use an image editing software like Photoshop and meticulously manipulate the data manually for each layer of the scan. These manual adjustments are prone to mistakes and can easily damage the underlying, physically correct data which we had devoted so much time and effort to capture correctly in the image acquisition phase. To mitigate these potential errors, colormass invested heavily in R&D to develop intelligent tiling algorithms to automate the tiling process. Run in the cloud on the colormass Platform, our algorithms not only significantly speed up the workflow but also guarantee the highest quality results.
Review and quality assurance. When tiling is complete, the 3D material is run through a standard set of tests. The results are visually inspected by colormass experts on different 3D objects in various illumination conditions to make sure that there are no undesired artifacts in the scans. This added step for quality assurance is necessary, as the fabric sample might have a slight fold or extruding piece of lint which was not originally apparent at the time of the region selection. If an artifact is found, the sample either has to be cleaned and ironed or a different part of it has to be scanned again. If everything went well, the material leaves the “Review” stage and gets set to “Completed”.
Cataloging. Every material goes through a series of steps: Draft → Info review → Scanning → Tiling → Review → Completed. This makes it easy to track the progress of the digitization process and organize the rest of the visual asset production around it.
When a material reaches the “Completed” state, a thumbnail is created for it in the form of a shader ball. This curved shape makes it easy to differentiate the various surfaces. Every material gets a name, an internal ID, and the article number issued by the manufacturer.
Once the above process is completed, you now have a high-quality data set that effectively describes and emulates how a specific textile reacts to light, creating a digital representation of the product's visual appearance. As explained in the sections above, this is stored in the form of PBR maps. These maps are considered as the raw output, which can be used to create different visuals. The next sections describe and show some of the most common use cases on the example of Maharam’s gorgeous Deconstructed Rose textile.
Flat tiles. Given the importance of flat tiles in the textile industry, they are placed in their own grouping. However, they are not really a category on their own; flat tiles belong both in the pictures and exports category described below. File tiles are simply a rendered image of the digital materials applied to a flat surface. What classifies them as a separate output type is the fact that these are special images in that they are tileable, meaning when placed next to each other both horizontally and vertically, they can cover an infinitely large surface without any visible seams.
%20-%2070%2C98cm%20x%2078%2C14cm%203k.jpg)
Marketing images. Probably the easiest to understand output are static marketing images. The digitized materials can be applied to any kind of surface in 3D. An upholstery fabric to the seat of a chair, a wallcovering to a wall, or wood grain to the frame of a bed. These items can then be rendered either on a white background or in a complete environment like a room of a house, a garden or a cruise ship. The results are static digital images, just like photographs taken by a camera. The interesting difference lies in the process of how these were created.
Configurators. Web based 3D technologies have come a long way and by now all modern browsers are capable of displaying 3D content using WebGL, without any additional plugins. Colormass developed its own rendering technology based on WebGL to showcase products in an interactive and highly realistic way. Instead of using a laborious and costly process of generating images from multiple views and in different variations, our 3D viewer allows customers to experience products in real time from any angle, in any variation. Visit the 3D configurator page to learn more.
Exports. The PBR data created from the material capture process is not only useful within the colormass ecosystem but can also be exported in different formats for further use by designers, architects or anyone else interested in producing high-quality visuals. Today there are more and more tools capable of using PBR materials. For example, CET Designer, pCon Planner and Revit now leverage PBR data and are widely used in the space planning industry. Blender, 3ds Max or Cinema4D also use PBR materials and are amongst the top tools to work with for excellent visual fidelity.
Each tool mentioned requires its own format and data conversion to use the material with the said software. Although the 3D industry is converging towards a common PBR format for 3D materials, it still has a long way to go before the interoperability between 3D software becomes a reality. Until then, the data has to be carefully converted for every single tool in order to ensure the highest visual quality possible. Luckily, the colormass Platform takes care of this, so you can be confident that your data can easily be accessed and used by architects, designers or 3D artists in their preferred visualization tools.
More often than not, textile manufacturers offer a very wide range of products including categories such as woven upholstery, leather, wallcoverings, acoustic materials, window coverings, rugs and digitally printed materials. The entire catalog easily amounts to thousands of SKUs. Although the goal should be to digitize all of the products using the newest PBR scanning technology, this often turns out to be too big of an undertaking at once. Hence, we at colormass normally recommend a two-stage transition process.
If you have little or no experience with 3D, it could be too risky to jump on the 3D bandwagon without understanding its implications on a smaller scale first. At the current moment there is likely an established process for creating the flat tiles of the textiles. This might be an in-house person who is photographing the flat samples and making them tileable using image processing tools like Photoshop. Alternatively, you may have one or multiple service providers whom you ship physical samples to, which they scan and process to deliver the digital files for you. In the initial stage of the transition process it is advisable to start upgrading this step for new products first. Instead of photographing or scanning (with a flat-bed scanner) the samples at a launch the regular way, start creating PBR scans. As described above, the flat tiles can be generated based on these scans. Even if the PBR maps are not used for anything else right from the beginning, a library of high-quality digital data starts building up which can be leveraged at a later stage. From here it becomes easy to start replacing further elements of the product launch process, for example the creation of images where the product is shown in use, e.g. applied to a furniture piece. As you learn more and more about the possibilities of 3D, new marketing tools like the configurator can be introduced to augment the success of the new product releases. The important piece of advice is to integrate the PBR scanning into the launch process early one.
Once the importance and the value of 3D has been proven for the new products, it is time to start considering the second stage of the transition: the digitization of the backlog. Although it can seem daunting to ship and digitize thousands of samples of existing products at first, it becomes much more manageable if organized the right way. First of all, the project can be set up to be accomplished over a longer period of time, for example over one, two or even three years. This does not only make the logistics easier, but the allocation of the budgets becomes more realistic as well. Another aspect to consider is the prioritization of the products. Starting with the bestsellers is almost always a good decision, working off the list to the products with lower sales volumes. Fabric ranges which are getting phased out can be skipped, which reduces the total amount of SKUs to be digitized as well. This way a seemingly huge undertaking can turn out to be a side project with a positive return on investment.
This blog article started out with a topic familiar to every textile manufacturer: flat tiles. It introduced the concept of physically based rendering (PBR) and showed a visual comparison why PBR scans deliver much higher quality outputs compared to simply capturing regular photographs of samples. Next, the steps of the digitization process were discussed in detail to support the argument of superiority for this next generation approach. Possibly the most interesting and also important part of the article followed, showcasing some of the possibilities these high quality scans open up: the creation of various visuals including flat tiles, application images and even interactive configurators. The last section suggested a two-stage transition process for textile manufacturers.
We at colormass are convinced that 3D is becoming an increasingly important part of marketing and sale of any tangible product. Companies who start investing in their 3D digital strategy today will be a step ahead of the competition, which will likely aid them in gaining a larger market share. One thing is for certain: the prevalence of 3D in the physical product space is inevitable, so wouldn’t you want to equip yourself for the disruption?
Want to receive more articles like this?