There are a million different recipes for baking a cake, but they have a lot in common. You'll usually want to gather ingredients, preheat the oven, make the batter, bake the cake and apply frosting. Sometimes maybe you'll want to start preheating the oven before you gathered the ingredients and that's okay. Some cakes don't have frostings, so the last step might not apply at all. The exact details of step number three, making the batter will be very different depending on whether you're making cupcakes, a vegan cake or a wedding cake. But overall, there is a general pipeline that most good baking recipes and visualization recipes tend to follow. Like baking any type of cake, making any type of visualization follows the same basic high-level steps. First, you have to explore the data to find out what's happening in it and to pick what pieces of it you may want to visualize. Next, you ingest the data into whatever software you're using. Then you design the scene which includes the appearance of your data set and in contextual elements and the treatments of features like cameras and lighting, and finally you generate images. A lot of these steps are like making the batter of a cake. The exact details will be different depending on what exact data set you're visualizing, the software you're using, whether you're making a single image or a video and a whole lot of other factors. Let's take a look at a more detailed workflow, often called a pipeline, an example recipe for creating a cinematic visualization video for a documentary. The first step is to acquire a data set. There's a lot of data out there and finding the right data set can take some trial and error, and making a connection with a data scientist who's capable of and willing to help. Next, you'll likely have to process the data in some way like by filtering it. A raw data set is likely to have a lot more information than you can reasonably show in a visualization, and it can have bad values and glitches so it's a good idea to tidy it in some way. Pre-visualization is at the core of exploring your data. This means getting a first look at your data in a quick and dirty way to familiarize yourself with what's going on in this particular data set. Here's a pre-visualization of a 3D data set of a coronal mass ejection on the surface of the sun in software called Party View. It can also be helpful to analyze your data mathematically, not just pictorially to see things like histograms, data ranges and other numerical information. Next, you'll likely have to do some more data processing to translate the data into a format that you're visualization software can understand, after which you can import your data into the software. There are lots of different options for software you can use to make your visualization. For cinematic visualization, you may want to use Hollywood film making tools like Maya, Blender or Houdini. Most of the examples we'll be showing you in this class will be in the software Houdini. For more traditional scientific visualization, a scientific tool like Visit, ParaView, VMD or YT maybe a better bet. Visualization tools like Tableau are meant for visualizing 2D relational data, so they're not appropriate for 3D scientific visualization. There'll be a lot of differences in your exact visualization pipeline depending on the tool you choose to use. See our course page for a non comprehensive list of software options. Some software costs money, some is free and some is open source. An open source piece of software means that the code used to create the software is distributed freely and anyone can download it or contribute to it. Many scientists like working with open source software because it means you can edit the code to do just about anything you want it to do, and you can benefit from other people's contributions. Now that you have your data loaded into the software we can start designing our scene. The next step is to create derivative data. Creating additional data based on the input data can make it more understandable to an audience. For example, you might release streamers into the wind field of a hurricane data set to watch how the wind carries them through the data. This gives you a better understanding of the wind than you would get by just visualizing it directly. Next, you'll want to bring additional data sets to the scene or design new elements to create a context for your data. Well, this visualization could just show the main data set of the earth, adding a milky way background of stars completes the picture. Next, we have to decide on a viewpoint position if we're creating a single image or a camera path if we're creating a video. A camera path is align in 3D that describes the path a virtual camera takes through a scene overtime. Now, we move on to designing, the coloring, transparency, material, and overall look of the data. A data set can look very different depending on how you choose to color it. Coloring isn't only important for aesthetic reasons, but can be used to highlight entirely different features in a data set. Once you're done designing your scene, it's time to create your images. The process of taking a virtual 3D scene and turning it into a 2D image is called rendering. It's quite similar to a photography, where you take the 3D world around you and flatten it into a 2D photograph, but it's not quite as fast as taking a picture. It can take anywhere from a fraction of a second to a day or more to render a single image depending on what exactly you're rendering and how. Once you have your rendered images the next step is to process them. You can combine multiple layers together and touch up your images in 2D to make them look nicer. A few software options that do this are Nuke, Premier or After Effects. The final step in this pipeline is to add annotations. This includes things like credits at the end of the video as well as relevant information on top of the video, Like a legend showing what each color means, how much time has passed and other labels. Your project pipeline will likely differ from this one in some ways. You may have more or less steps or you might do them in a slightly different order. Something that will almost certainly happen is that the pipeline won't just be a one-way street. Making a visualization is an iterative process. After you render a version of your visualization, you're likely to find things that you'll want to improve or change or things that you've done wrong so you'll have to go back and redo an earlier step. This is an expected part of the process. It's important to remember while you're working through the pipeline to take care of yourself. It's not uncommon to hear about repetitive stress injury and carpal tunnel among people who spend a lot of time working with computers like those doing graphics or programming. An ergonomic mouse and keyboard can help, and so does maintaining good posture. Spending a whole day looking at a screen one to two feet in front of you is straining on your eyes. You can actually get special screen glasses to help with eyestrain. These are mine. Decent hardware with the highest screen refresh rate will also help. Give your eyes a break by taking some time to look outside often to the distance to relax your eye muscles. When working on a real project in industry or academia, you're not likely to need to take on the whole pipeline by yourself. Plenty of people specialize in just pre-visualization or just camera design or just rendering. But it's important to understand the whole pipeline and how the pieces fit together in order to work efficiently and effectively. Fun fact, in Hollywood studios all of these parts of the pipeline tend to move in parallel, if not on the same movie then on different movies at the same time. There's actually another piece of the pipeline, that's the pipeline itself. People in charge of this work with pipeline management software like Shotgun or F track to keep everything in the machine moving smoothly.
