Welcome back dear readers!
It’s time for us to leave our field work behind, and get comfortably seated in front of our computers – since from now on, we are going to use them a lot. Magic is about to happen! Having taken care of our pictures as described in the previous post, we’ll change our camera and tripod for software such as Agisoft Photoscan, zBrush or Maya. But even though we are moving into another stage of our project, we should keep in mind that much of what we’ll be doing from now on relies on our previous work – therefore, having a good foundation is key to achieving good results.
So where do we start? Now that we have our raw data from the images, we should take that into the computer. It’s a good practice to have everything organized and keep a neat folder structure, where we store our pictures and meshes, since we’ll end up having many different files which in turn will be part of different stages of the project. Don’t underestimate the importance of this step, since we’ll be working with many file formats of the same images, many computer programs, and at many different stages – which can make it very difficult to locate a specific file.
And talking about having things well organized, under these lines of text I’m going to paste a sequence of pictures that describe the following steps we are going to follow. This post can be a bit word-heavy, so I’ll give you a clear view of what we’ll be doing now and you can read later in more detail about each of those steps.
So now that you’ve taken a look at the images above, it’s time to introduce the first program in our pipeline: Agisoft Photoscan! We can use other programs as well, such as Reality Capture, but today we’ll cover the basics of the first one. We’ll use it to build a cloud made out of points – commonly referred as a point cloud – which will later turn into our highly detailed mesh. This is a piece of software that we’ll be coming back to a lot throughout the rest of the project, and it will be our main source for creating the colour textures. Let’s take a look at it.
The first step is importing our images into the program. To do so, we’ll go to the ‘Workflow’ tab, and once in there we’ll clik Add Photos, where we can import either the shots we want or a whole folder which contains them all.
After that, Agisoft will need to align the cameras in order to ‘understand’ the scene. By doing this, the software will create a point cloud of the analysed data, which we will later turn into our mesh. There are different presets here, with quality settings that range between low and high. Choosing between them is a matter of knowing your hardware, your time constraints and the level of fidelity that you want to aim for. I’ve only used Agisoft on my computer this far, but using the lowest configuration could take up to 30 minutes, whilst using the best one might clock in the 3 hours. Of course, this always depends on your machine at home and the scene itself, but do note that there’s a noticeable difference between the settings.
Once that’s done, we should focus on selecting the area that we want to work with. This step comes in handy in almost every scene, since the program always generates a point cloud bigger than what we need. Therefore, prior to any further work we do, we should establish the boundaries of our desired model, so that the program spends more time processing the data that will actually be part of our scene. To do this, we can go to the main toolbar and select any of the three icons to the right of the selection tool – the arrow -, which will let us select any undesired points and delete them.
After that’s done, there’s a couple of things that we need to do before building our mesh. First, we should align the cameras a little bit more – we can do this by specifying different points in the images that appear throughout the rest of the lot. To do so, we just need to pick the selection tool, double click in any of the images, and once in there, create a marker by right clicking any desired point in the shot. We can then go to another image, and click again on the same spot. The program will actually help us with this part of the process, since it will try to identify by itself the points that you’ve created.
The next item in our ‘things to do list’ before we can actually create geometry is to generate a dense point cloud, which is the last thing that we’ll cover in this post – we’ll leave the creation of the geometry for the next one. Doing so is very easy, but it relies on our previous processing of the pictures. And this is a path that we should tread carefully, since building the dense point cloud is one of the most time consuming steps. The quality of the build will be the result of the accuracy of our previous work, and since our computers will take a long time to process this information, we shouldn’t have disregarded any of the previous operations.
In order to create this new cloud of points, we should just head to the workflow tab and, once in there, click on ‘generate a dense point cloud’. Agisoft will present us with various options with regards of what quality we are aiming for: lowest, low, medium, high and highest. There are some other parameters as well, and I’ll leave you a link to the program’s manual in case you want more info on this topic, but I want to cover the differences in time that choosing between the general options presents you. On one hand, the ‘lowest’ and ‘low’ settings will require your computer to stay working for a couple of hours – again depending on your hardware -, but even on high end pc’s it can take up to 4 or 6 hours. Medium quality, the one I usually aim for, can set you back 24 hours. A whole day just spent on building a dense point cloud. And that’s why I choose it, because when we aim for a ‘high’ or ‘highest’ settings we can end up wainting for 3 days.
And that will be it for the moment! Having reached this step, I think it’s a good moment to pause for a little bit and let our computers work for a little bit more… maybe a lot more, it just depends on what settings you’ve selected! We’ll continue on the next post talking about what comes after we have our dense point cloud computed. Until then, I hope you found this useful!
Read you soon!
————-
Brais Brenlla 