As the progress continues, this is the next step to building autonomous robotic systems.
From what I understand so far there are two ways to create computer vision. one is a Vision Transformer network, which I haven’t had a chance to learn yet , and the other is a convolutional neural network , which is what I’m doing first
In order to build a convolutional neural net work you need a convolution and convolution needs edge detection (need may not be accurate), because as many articles say you send a convolution into your neural network for it to predict its results. A convolution is used to create a combination of features. Some of those features include edge detection, and I had not done that yet; not from scratch at least, so I taught myself how to do it.
At first, I wrote my convolution using the discrete Fourier transform, and then I ended up just using a dot product. Eventually I might go back to my discrete Fourier transform, which was so much fun to learn but this is working
This is in pure Vanilla Javascript, and is awesome. There are still sophistication that need to be made however, right now I can send in a simple image. My software can detect the lines and edges, using a Sobel kernel and re-create the image from its edges. That is exciting. That is the first step to Computer vision as I understand it.