This is an introduction guide to shooting Product references. This will usually go alongside 3D scanning and is aimed to capture product geometry.
Aims for Shooting:
1. Flat views for a 360 view of the product
2. Detailed views of the product.
3. Measurements of detailed components (especially components the scan will not measure)
4. Dimensions taking and sketching
5. Texture References with Colour Card- Please refer to Texture Capture 101
STEP 1: Flat views
You are aiming to shoot flat views of the product to provide a full idea of product.
Long Lens required.
Shots required:
3/4 View
Front
Side Left
Side Right
Back
Top
Bottom
STEP 2: Detailed shots
Aim is to pick up details of the product. Such examples are:
Arms, Creases, Stitching, Cushions, Buttons, Handles, Hinges, Glides ect
Anything that required a zoomed-in shot, and that may be missed by the flat shots.
STEP 3: Measure the detail
Use a tape measure to give real distance and scale to detailed parts of the product. It is important that the focus is on things that the 3D scanner will not measure accurately.
Examples are:
Any Chrome features.
Any dark areas of the product.
Anything towards the base/ underneath the product such as feet.
Tape measure direction will depend completely on the general shape of the product.
STEP 4: Dimensions Overall
The last step, is to take a few measurements of the product, this is to eliminate any debate on the overall size when the scan is perhaps being scaled. It will also be useful for recognizing which versions of products they are such as mattresses being Double or King.
This method is typically done by doing a quick sketch and measuring a few key dimensions, see example here:
STEP 5: Colour Card
Please refer to Texture Capture 101
Complete. Next steps are to take references of the material which will be covered in a separate section.
Recap and summary:
STEP 1- Main views typically at least 7 shot. Long lens.
STEP 2- Details. Zoom in on parts and sections of intricacy.
STEP 3- Tape Measure. To measure parts the scanner will not scan.
STEP 4- Sketch and rough dimensions.
STEP 5- Refer to Texture Capture 101
1. Different models require different references - we can categorise the models based on their function, size, materials, etc.
The main categories are: Large Appliances, Small Appliances, Furniture, Large Props, Small Props, Lights, Plants, Decor. Be aware of their specific requirements before starting to make the references. Further information about the different categories will be provided below.
2. Try to shoot flat views of the product from all sides: front, back, side, top, bottom and several pictures from different angles. Carefully examine the object and make pictures of the smaller details. Avoid camera distortion by using long lens. For in-depth info check "STEP 1: Flat views".
3. Try to ensure good lightning when taking pictures. Ideally it needs to be soft, even, without big hot spots. If the place is too dark use your own light source.
4. Aim to provide the sharpest possible images. Check if you have the correct focus distance especially when you try to capture small details, inscriptions, labels, fabric textures, etc.
5. When you take dimension pictures with measuring tape make sure that the object and the measuring tape sit at an equal distance from the camera. If not we receive inaccurate due to the perspective distortion.
6.For oval shapes use caliper. Due to the perspective distortion is is very inaccurate to measure circular objects with measuring tape.
7. Try to shoot objects in their original position - this is especially important for soft objects that are affected a lot from the gravity(blinds and curtains for example).If not we might be missing important details.
Usually large appliances are machines used for routine housekeeping tasks such as cooking, washing laundry, or food preservation. Their common trait is their big size and they are difficult for transport. So most likely the references needs to be shot in an unfamiliar environment. Pay attention to the lightning, choose a good spot to take make the pictures. Lens distortion can be a problem, use long lens if possible. Aim to shoot flat views of the product from all sides: front, back, side, top, bottom(if possible) and several pictures from different angles. Use steel measuring tape to cover most important dimensions.
Large appliances usually have a lot of small details, buttons, inscriptions and labels that need to be captured separately. Try to make clear pictures for all of them. You might need to adjust your camera to capture everything as sharp as possible.If there are opening parts make several more pictures in different positions.
A small appliance, small domestic appliance, or small electrics are portable or semi-portable machines, generally used on table-tops, counter-tops, or other platforms, to accomplish a household task. Examples include microwave ovens, toasters, humidifiers,food processors, coffeemakers and others. They are much easier to move and place in proper position for shooting. Aim to shoot flat views of the product from all sides: front, back, side, top, bottom and several pictures from different angles. Sometimes those objects have very complicated geometry, be sure to capture every curve, detail, inscription, button, etc. If there are movable parts make several pictures in different positions. Use steel measuring tape or caliper, depending on the form of the object.
Probably the most difficult objects to make proper references. They can be divided to two big groups - hard and soft bodies. For the hard surface models aim to make flat views of the product from all sides: front, back, side, top, bottom(if possible) and several pictures from different angles. Use steel measuring tape to cover most important dimensions. Check if the light is enough, if not add a light source, or move the item to a better place. Some of the objects will be quite big, make sure to use long lens for the pictures. This is very important for models without right angles and for models which are defined by big curves rather than lines. For those we can receive very altered look for the object if shot incorrectly. After we have the main references make some close up pictures for handles, hinges, buttons, seams, etc.
For the soft surface models again we follow the basic rules - flat view from all sides,proper lightning, measuring of the main dimensions. Those objects usually have very complicated form, so making the perfect picture is impossible. The best we can do is to analyse the form and try to capture the most important curves, bulges and creases. Pay attention to smaller details like ridges, folds, zippers - they are very important for the final look of the model. Make sure to capture those properly. In hindsight the best way to reference those kind of models is to use scan data for the main shape and the pictures to serve as additional source of information.
Furthermore references can be provided by the client. Those can be in form of drawings or simplified 3D models. This is especially important for mechanical details like lifting mechanisms, joint plates, bolts, hinges, etc.
Overall the information can be gathered from a lot of different places depending on the model, the client, the available time, etc. But the simple rule for those kind of models is: The more information available-the better and more accurate model.
Props are mostly used for scene fillers. Depending on the client we may or may not have physical samples to make our own references. If we do, these models should be considered "client products" and we need to follow all directions from the above categories. There are a lot of versatility with these models, but once again the aim is to provide flat views from all angles without too much distortion and with proper lightning. If the models have unusual materials we need to make additional references about them also. If we don't have references from the client usually we search for them in Internet. Those models usually have limited amount of information and are used for scene fillers. The most important task for those models is to have accurate geometry and the textures to be as close as possible to the references. If exact textures can't be found the goal is the materials to have similar feel and look as the original.
In theory "LIGHTS" could be included in "PROPS", because everything we said above is also true for the category. The one big difference is the lights have two modes. When we have a physical sample we should make additional pictures with the "lights ON". Every light has it's own properties (type, intensity, temperature, shape), so if we want to represent accurately those models the information could be found on Internet or be requested from the client. The best way to replicate the light source is to input physically correct values. Another tricky part is the lampshade. A lot of times it is translucent, we aim to make it as close as possible to the reference.
The final and approved model should have two renders: "lights ON" and "lights OFF".
When we make references for "BLINDS AND CURTAINS" ideally we want to place them in their original position. This is more time consuming and requires special setup, but it is the only way we can receive full and accurate representation of the model.
We also need to make some closeup pictures, so we can fully understand the structure of the model - elastic cords, seams, folds, etc. Since those models are created via simulation the more information we have, the more accurate the final result will be.
Another very important feature of each curtain is its bottom part. Depending on if the item sits on the floor or hangs in the air we receive completely different result. This information should be requested from the client before the actual modelling is started.
In many cases we need to add labels or inscriptions to a product the client requested. In those cases the best outcome for us is to receive a vector file with those labels and the fonts needed. Since some fonts are designed specifically for the client he is the only one to have them. If not-the solution is to search for a similar fonts, but the result will not be 100% accurate.
Another way to fix this problem is to make proper reference pictures. Since most labels sit on rounded shapes(bottles, jars, etc), one front picture is not enough. The best course of action in this case is to remove the label, flatten it and then shoot it.
Now let's look at some references and try to analyse them.
Here we have a very good set of references. We have the main flat views, most important dimensions and some perspective pictures. Movable parts have additional pictures, there is even a picture of the object in working condition.
In this set of reference on first glance we have something similar to Example 1 - good flat images, a lot of dimensions and some additional pictures. But there are two main problems. We are missing the overall dimensions, which are the most important. Those are critical if we want to make accurate model. In this case we have a good pictures without distortion, so we can make correct geometry directly from the picture, without looking at the smaller dimensions. The second problem is that we have too many small measurements, which costed a lot of time but don't serve any purpose(most likely there are also not very accurate because of the small size and lack of caliper). In hindsight we have a good set of pictures, but we can optimise the way the dimensions are presented.
In this example we will meet a very common problem - replicating complicated material from reference. The task is to make this table, requested from a certain client. On the first row we have refs from the Internet, on the second - our refs and on the third - attempt to reference the tabletop.
On first glance this is a pretty easy model, but we face some unpleasant and time consuming issues. In terms of referencing the dimensions we see some missed dimensions, when we have dims it's not very clear which parts are measured. It is still possible to make fairly accurate model, but it will take more time and sometimes there can be some issues with the proportions. The bigger problem comes when we try to replicate the wood texture. Since this is a "Client Product" we need to make our wood material as close as possible to the reference. Finding similar pattern in Internet is almost impossible, so we need to rely completely on our own references. But there are several issues. We don't have overall flat picture of the tabletop, there are only small patches. Each of them has different lightning, different size and there is light gradient in them. So it is impossible to make a full size texture out of them. The solution is to make one frontal picture of the entire tabletop with even lightning. This should be the most important shot of the entire session since the geometry is fairly simple and easy to make.
For this type of models usually we have scanned data. If we don't the main dimension needs to be measured. The most important dimension here is the HEIGHT, since the LENGTH and WIDTH are bound to several bed sizes.
The biggest problem with these models is replicating their materials. They are built from different fabrics and each of them has complicated shape and properties. To extract all data needed from pictures made in non-controlled environment is impossible. Uneven surface, bad lightning, blurred photos and reflectiveness of the materials all contribute to this issue. There are several ways to solve this problem. The best and most accurate way is to scan the fabrics. For that the client should provide flat pieces of the cloth. From the scanned data we can extract tileable texture maps. The other way to make those fabrics is to combine pictures with custom made bump, normal and displacement maps. Usually this method is much more difficult, time consuming and inaccurate. Sometimes there are additional shapes imbedded in the base fabric(like the stylised flowers-ref picture above). Tracing those is very time consuming and inaccurate. Ideal scenario is to ask the client for those patterns,logos and everything connected to the brand because they already have them designed.