PROCESS SO FAR¶ ↑

2015¶ ↑

• 20 June - Nick suggested computer vision schedule

• 21-25 June - Attempt to install ROS Indigo on SnowLeopard (which corrupted 32-bit OS set by MacBook EFI)

• 27 June - Realised that need to reinstall SnowLeopard

• 6 July - Reinstalled Rails/GitHub sys (after rEFIt Ubuntu/OSX)

• 7-11 July - Help Iain move out from Kinrara

• 14 Aug - Sent email to John saying “no” to 80th party

• 30 Sept - Download Computer vison models book

• 14 Oct - Download Steffen Nissen's MSc thesis, FANN

• Winter - Study Bayesian stats

2016¶ ↑

• 23 April - Not attend John's 80th party

• 3 June - Not attend batch 20yr reunion at CTCRM

• June - Finish reading BarronMalik paper

• June - Finish LSD-SLAM paper

• 16 June - Install and run lsd_slam_viewer

• 18 Nov - Study OpenGL (having corrupted my Ubuntu partition in upgrade to 16.04-Xenial Xerus)

OPENGL¶ ↑

OpenGL Shading Language¶ ↑

Control of graphics pipeline without ARB assembly language or hardware-specific langs

• en.wikipedia.org/wiki/OpenGL_Shading_Language

OPENCV¶ ↑

docs.opencv.org/2.4/

• Alternative to Matlab

  www.learnopencv.com/opencv-c-vs-python-vs-matlab-for-computer-vision/

Python¶ ↑

(25/1/17)

github.com/opencv/opencv/tree/2.4/modules/python/src2

Check version:

$ python
>>> import cv2
>>> cv2.__version__
'3.1.0-dev'

$ env
PYTHONPATH=/opt/ros/kinetic/lib/python2.7/dist-packages

EIGENFACES¶ ↑

“Eigenfaces is the name given to a set of eigenvectors when they are used in the computer vision problem of human face recognition”

Vector¶ ↑

A list of data items like a 1-D array

Arrays¶ ↑

“We can think of a vector as a list that has one dimension. It is a row of data. An array is a list that is arranged in multiple dimensions. A two-dimensional array is a vector of vectors that are all of the same length.”

Practical implementation¶ ↑

www.youtube.com/watch?v=SaEmG4wcFfg

• Prepare set of training images in single matrix T, where each col is an image (a vector)

• Avg image calculated and subtracted from each original image

• Calc eigenvectors and eigenvalues of covariance matrix, S

  Eigenvectors of covariance matrix are called eigenfaces

  They are directions in which images diff from mean

• Create threshold E on total variance

  k is smallest number satisfying:

  n(lambda-1 + lambda-2 + … + lambda-k) / variance > E

If working with 100 x 100 image then obtain 10,000 eigenvectors

but most faces can be ID'd using projection on 100-150 eigenfaces (so reduce computation)