GP is a kind of programming that will generate code for you at compile time.

The core is to seperate algorithms from the data structure, and we do some abstraction over a data structure,

and then work with that abstraction instead of the actually type.

 

 

template errors are insanely big.

Vector is a template class.

 

 

 

 

 

 

 

 

 

http://en.cppreference.com/w/cpp/language/class_template

http://www.drdobbs.com/moving-templates-out-of-header-files/184403420

 

method 1:(Declare and define in header files) if declare and define both in header file, then we dont need to link files. (header on the library)

method 2:(Declare in NAME.h file) hpp is just a habit of naming such a file, and #include just means add those content of hpp file into header file

method3 : /

 

 

 

 

 

func .find in map object

 line 15 is an efficient way to check if 3 in map

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  

line 3-8: the first template

line 10-11: the 2nd template

 

question: what's the meaning of "::" in line 15? 

 

 

free BSD license:

free for commercial or non-commercial

 

 

 

 

 

 

 

 

 exr is an open format to store data, and opencv supports it

 

 

 

 

My implementation:

 

 Gotcha!

 

 

 

 

 

 

 

line3 type: unsigned char 3 channels

and  line 11 is: float 1 channel

 

 

 

now DNN is a new and better descriptor but it's slower.

 

 

 

build a KD tree and query for neighbor, much faster than brute force search 

 

 

 

SIFT is non free, and is in libopencv-nonfree