The next action step in my research is to try out machine learning, specifically an existing CNN (convolutional neural net) and a new LSTM (long-short-term memory) RNN (recurrent neural net), on embroidery patterns. Well, on basically images turned into serial vector art that will be translated into embroidery file formats.
Now, I technically failed my machine learning class because of disagreements with the professor and some slight sexism, but I have a basic understanding of statistics and simple machine learning concepts (and the weka toolkit). CNNs and RNNs were out of that field of study, so the specifics of how to run data and build one is kind of a mystery. I can’t find much of a tutorial, but I do have a friend helping me, Adam Summerville. His first linked reading for this was THIS awesome blog post that goes into a lot of depth, but still seems pretty readable with lots of diagrams and pictures. These are the notes I stepped away with:
Step (-1) at the bottom of that image is setting up Ubuntu, graphics drivers, and figuring out if I’m working with AMD or NVIDIA (pretty sure the latter, thankfully). Based on the graphics card, I figure out which NN drivers will work for me (I believe CUDA suggested 7.0 and Torch, which also means lua). I’ll update this page as I make progress. Installing an OS from a 2.0 USB takes a while…
Also THIS will be helpful as a Newbie Torch tutorial.