I’m currently working on a research product into using neural networks for generative design. To train the neural network, I want to create a data set that matches a 3D render with some input parameters for the design made using dynamo.
Is there a way I can automate feeding dynamo a list of parameters and produce a render, without having to go through each permutation individually?
Take a look at this, it might help you!
Thanks for the link, that will be really useful for this project. It doesn’t quite solve my current problem sadly because it seems to address the optimisation stage.
I think my initial explanation was poor.
I have already created a node that takes a number of inputs and generates a geometry. What I want to do is generate a matching render image for each set of inputs, so that it can be scored by a human on it’s beauty. I could do this by hand, but I wonder if there is a way I could generate an arbitrarily large set of inputs and their matching renders.
Sounds like you’re making something like Project Fractal. Would love to see what you’re developing and how you see it could be applied.
I am not aware of a way to export the backround preview as an image file. Exporting an stl is doable and may need to be considered of yi are doing things in 3D space.
Alternatively maybe you could project lines onto a plane, thicken them, read the intersect values between the thickened lines and a given point array on the plane, and set color values based on that, effectively making a bitmap. Look into the Image.FromPixels node and the Image.WriteToFile node to help with this.
@Keith_Alfaro FYI -
@H.rose.2 I believe there is a method for this exposed in the API in Dynamo 2.x but I don’t think there is an OOTB node, I think Keith may have worked on a package for this at some point, but it requires using the daily builds if it still exists.
Thanks for the help guys. I think my plan is to export the stl files using meshtoolbox, and write a dedicated program that views the stl files and allows the user to score it. Since I already needed to write a tool to combine the scores and inputs, this shouldn’t complicate things too much.