Index

 

Gesture Recognition with a Neural Net

I used a feed-forward multilayer neural network to do gesture recognition of Wiimote motions. This is the paper I wrote about it: Gesture Recognition


Gerster Waves & Boids (a.k.a. FishTale)
CpSc 426 Project w/Sahar Sajadieh

For comments or questions, please go to my Blogger page: Blogger

Downloads:

Video: Gerstner Waves & Boids

YouTube: Gerstner Waves & Boids

Win32 Binaries: FishTaleWin32.rar*

Src: FishTaleSrc.rar

*I'll be updating the binaries with a few more options soon. This should allow it to run smoother on lower end PCs.

Concept:

  • Deep ocean scene with Gerstner modeled deep ocean (infinite depth) waves and boiding flocks of fish.
  • Fish have a slight attractive force to the camera, and repulsive force from the water plane.
  • Freemoving FPS-style camera.
  • Spheremapped water texture.
  • Skybox is used to create background.
  • Fog to create sense of depth.
  • Single directional light in -y direction.
  • Models in OBJ format are supported.
  • Textures can be any format supported by DevIL.

Gerstner Wave Algorithm:[Tessendorf 01]

  • x0 - Undisturbed horizontal position on surface.
  • y0 - Undisturbed vertical position on surface.
  • A - Amplitude of wave.
  • k - Wave vector.


  • x = x0 - (k / |k|) A sin(k . x0 - wt)
  • y = A cos(k . x0 - wt)
  • Multiple waves can be summed.


See: RenderableGlWater.h & RenderableGlWater.cpp

Boiding Algorithm:

  • d - Distance.
  • x0 - Average position amongst neighbours.
  • v0 - Average velocity amongst neighbours.


  • Separation force - Repulses boids from neighbours based on 1/d
  • Cohesion force - Attracts boids to x0 based on d^2.
  • Alignment force - Aligns boids in direction of v0.


See: Boider.h & Boider.cpp

Physics:

  • Euler integration.

Libraries: