Support for data science, machine learning, and more - within Kojo.
A previous cut of kojo-ai is available at https://github.com/litan/kojo-ai. This is the next cut, based on the official Tensorflow-Java API.
- Neural Networks (via Tensorflow-Java).
- Plotting (via XCharts).
- Graphs (Bfs, Dfs, A* search, etc).
- Neural Style Transfer (NST) (via Tensorflow-Java and Arbitrary Style Transfer).
- Object Detection (via Tensorflow-Java).
To use Kojo-AI within Kojo, just use a published release, or copy artifacts from this repo into the Kojo libk directory as per the following steps:
sbt buildDist
mv ~/.kojo/lite/libk ~/.kojo/lite/libk.bak
cp -var dist ~/.kojo/lite/libk
An example showing NST in action is given below:
// #include /nst.kojo
cleari()
clearOutput()
val alpha = 0.8f
val fltr1 = new NeuralStyleFilter("/path/to/savedmodel/", "/path/to/style.jpg", alpha)
// val fltr2 = new NeuralStyleFilter("/path/to/savedmodel/", "/path/to/style2.jpg", alpha)
val drawing = Picture {
setPenColor(cm.gray)
var clr = cm.rgba(255, 0, 0, 127) // start with a semi transparent red color
setPenThickness(8)
repeat(18) {
setFillColor(clr)
repeat(5) {
forward(100)
right(72)
}
clr = clr.spin(360 / 18) // change color hue
right(360 / 18)
}
}
val pic = effect(fltr1) -> drawing
// or apply two filters
// val pic = effect(fltr1) * effect(fltr2) -> drawing
draw(pic)Original Kojo Drawing:
Drawing with one NST filter applied:
Drawing with two NST filters applied:
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent