This repo contains experiments for learning more about computer vision and algorithms.
It is a personal quest for learning techniques used with computer vision, experiment and create new algoritms.
Background
As a platform to experiment on I have chosed Android because:
- It provides a parallel computing API called RenderScript
- The platform itself is a device with integrated sensors (camera, microphone, positioning)
- Most modern high-end phones provide enough computing power (as far as I can see)
Alternate platforms include CUDA but I do not have the hardware available everywhere.
Lastly, there exist libraries like OpenCV for performing many computer vision operations with ready made implementations. Currently I want to focus on just the bare implementation without a library for the following reasons:
- I want to get hands-on experience developing algoritms and eventually know enough to develop my own
- I want to get hands-on experience with parallel computing with kernels as used in renderscript / CUDA
- I am currently unfamiliar with many of the mathematical concepts like matrixes, eigen-values, neural networks. Learning OpenCV is too steep
APK - See https://github.com/udevnl/computervision/releases
Videos (click to watch on YouTube)
Vector based edge detection
Interest Point Detection
Particle forces simulation
