Pandemic Pop-Up - Image Capture Workflow

Pandemic Pop-Up – Image Capture Workflow for 3D Modelling

Background (for those interested in the process – skip if not interested)

Photogrammetry is the process I use to create these types of models. It involves taking numerous photographs of an object from many angle. These images are then fed into the model building software. Essentially what the software does is take a photograph, identify features that it believes to be unique in the photo and then try to find those same features in the other photos in the set. Once it has identified the same features in multiple photographs it can calculate the change in the relative distances between feature points between photographs. With this information the program can determine where the cameras were in space relative to the object when the photographs were taken and build a cloud of points that represents the shape of the object. From there its basically a matter of playing connect the dots with the point cloud to create a surface model and projecting a photographic texture onto it.

(except this part – read it) Models can be made using either still photos taken with pretty much any digital camera are by isolating frames out of high-definition video. For the Good Mischief proof-of-concept model I isolated frames from 10 minutes of 4K video shot with a Samsung Galaxy S7 phone. I have posted part of that video on YouTube to give a sense of how to capture the angles.

Before You Begin

If you are setting up a display specifically to try this there are a number of things you can do to help increase model quality.

Shiny Objects Bad - Unfortunately very shiny reflective, or transparent objects are very difficult to model using this technique, this is why the glass shelves in the Good Mischief model look so weird.

Thin Objects Tricky – Very thin objects can be difficult, part of why the little devil Halloween decoration did not model well. Fur, hair, and feathers can be problematic for this same reason. This can be countered to a degree if these features are up against a surface, like a wall or box. If they are free standing they are likely to be left out of the model, like the unicorn’s horn. Though if I had done cloe-ups specifically on the unicorn the horn would have modelled fine.

Strong Shadows – Strong shadows and uneven lighting can make it difficult to model the surface in the shadows. Try to make the lighting as even as possible.

Single color / Lack of Features – Large expanses of a single color make it difficult for the software to find features to match. Smooth surfaces without a lot of complexity also experience this problem. For example, a smooth white vase with a shiny glaze sitting on top of a lite blue box in front lite blue wall would be very difficult. The key is to try to give the software as much detail to work with as possible.

If you are working with an existing retail display, try to use a portion of your display that does not violate to many of the above considerations. That said, the final photographic texture can hide a lot of the sins of a subpar model, and right now this is all about trying to make the best of a bad situation.

Taking Your Photos (most smart phone photos should be fine)

If you are shooting still photos you will need to take somewhere between 100-400 photographs. Generally, the more photos the better the overall model quality, though in the interest of time and trying to get as many of these models made as possible I’d say try to keep it around 200. It is important that there be an approximate overlap of 30-50% between any two adjacent photos. The goal is to get as many photos of all the objects in your scene, from as many angle as possible, with a significant amount of overlap between images.

Still photos work much better, but sometimes video can be easier especially if you are in a hurry.

With video I have found that it really has to be 4K video, 1080p just really doesn’t record enough detail. With video the same concepts apply, scan slowly over the surface capturing it from as many as you possibly can. While just a few minutes of video should be all that is necessary, if it’s a situation where I may not be able to return to record again I will often shoot a full ten minutes (the limit my phone puts on 4K videos).

Once the Image Set is Taken

Once the image set or video is taken you will need to get them off of your phone or camera and place them in the Dropbox folder I’ll send you a link to. From there I can start working on putting the 3D model together and getting it ready to post on Sketchfab. Once the rough model is done I will post it in a private mode that will allow you to review it and we can work on how it should be annotated and what the description should include.

These are intended to just be one more, kind of interesting way, to get your work and wares out in from of more eyes during this difficult time and hopefully direct more traffic toward your online sales platform.

I am lucky enough that my day job allows me to work from home during this, I’ve chosen to largely put my own artistic pursuits on hold for now so that I can spend my free time doing this. However, since I am still working that means that in all likelihood I will only be able to get about one new “Pandemic Pop-Up” done per day once I get the workflow ironed out. If this is something you want to try please keep that in mind and be patient.

Stay safe, and let’s all keep looking out for each other.

-Ian