Training a CNN with openCv using C++
Hi,
I have basically a set of 2D Points. So, lets say, I consider a big 200x200 Matrix and initialise it with Zeros. Now,I scale all of my 2D Points such that its x,y coordinates lie in the range of [0,200] (ofcourse,Integers). So, now,I fill my 200x200 Matrix with One's,wherever I have my point coordinates. So,My matrix is a collection of Zero's and One's.
Now,I would like to train a very simple Convolution neural network.
Input: Matrix of size 200x200
Output: Labels as Integers (0,1,2,3).
Once CNN is modelled and trained, I would like to use it for classification. Please help!
@berak: Please do help!
"Training a CNN with openCv" it is not possible with opencv
@Programming_Enthusiast -- you heard it ;(
but if you want to play with cnn's there's e.g. tiny-dnn , darknet or even google's colab, on the web
@berak: Do you recommend using tensorflow? I tried a lot, but I did not find any kind of help,for implementing CNN in C++.Please do help for my case.
well if you want tf, i'd recommend the python version on colab (at least there's docs/tutorials for that)
Ok :(..Please atleast provide some sample code for python please! I would try out with my data..
I tried this https://github.com/LaurentBerger/Caff... (C++ caffe ) or http://perso.univ-lemans.fr/~berger/C... (tf and python)
@LBerger: Thanks Mate! I guess,they are considering training the data directly from Images right?
Lenet might be indeed an easy start for your problrm
@berak: Please do provide some sample code for Lenet. I would be glad,if you can clearly explain,how I can use !