Vision Union is a collection of computer vision workers that becomes more powerful together.
The 'workers' are image processing and 2D convolutional neural network building blocks, implemented as High-Performance C# jobs.
This project is a prototype and proof of concept, which i'm unlikely to come back to.
Look in /Assets/Runtime/Scripts for the code that actually implements CV stuff
Efficient runtime inference / serving of existing computer vision models on the cpu, by leveraging Burst and the job system.
Utilities such as RGB -> grayscale conversion, integral image calculation.
2D spatial convolutions across multiple channels. TODO - depthwise convolutions
Max2D & Average2D
ReLu & ReLu6, TODO - softmax