Computer Vision Using Ruby and libJIT
summary
Implements a Ruby vision library called Hornet that uses Ruby + libJIT to compile elementwise array ops into machine code, trades a 30× VM deficit down to sometimes only ~6× slower than C++, and falls back to tiny C++ generators or FFTW where needed. Builds a real-time augmented reality marker tracker using connected components, gradient histograms, line fitting and SVD homography estimation, demos working 3D overlay on camera video.
AI-written summary · about
chapters (26)
- 0:00Intro
- 1:33Examples
- 2:10Presentation Software
- 3:18I Robot
- 3:31I do it in C
- 4:10I do it in Ruby
- 4:24I use the approach
- 5:04The core of the library
- 6:45Generic integer classes
- 7:36Static arrays
- 8:41Arrays
- 11:22Speed up
- 11:34Reflection
- 12:13Limitations
- 12:29JIT Integration
- 13:20JIT Variables
- 14:23JIT Compiler
- 15:19JIT Loop
- 15:35JIT Illustration
- 16:08Performance
- 16:47Example
- 24:16Pinhole Camera Model
- 25:20Intrinsic Camera Matrix
- 31:30Source Code
- 32:41Demo
- 34:06Questions
related talks
26mDetecting and classifying object images using ruby
41mThe Future of Computer Vision: How Two Rubyists Are Changing The World
36mRuby is For Fun and Robots
23mBuilding C Extensions in Ruby
20mComputer Vision at the Edge with OpenVINO
31mExploring Real-time Computer Vision Using ActionCable
36mSome Assembly Required
33mComputer Vision, Not Just for Breakfast Anymore