1
$\begingroup$

Hello I am new to Applied ML and trying to solve a problem where I have given several images with few patches in it and masked images where these patches are classified/extracted as training data.Given that i need to use ML to train network such that It can generate those masked images for test data set. I would like to know which technique will be best suitable in Both the cases where I want to use Python extended libraries like Scikit or Using just Numpy and Scipy. Thanks in advance

$\endgroup$
  • $\begingroup$ The usual meaning of "several" is "about 5". Do you really have less than 10 training images? Could you give a little more detail, and if it is OK to share a typical original plus masked image it might help to choose an algorithm. If you really have just 5 training images, then you are not going to be able to train a classifier from scratch. $\endgroup$ – Neil Slater Jan 5 '17 at 11:13
  • $\begingroup$ I have around 164 images each of size 128 x 128. Train data contains images similar to this sweethaven02.com/MDSeries/MD0852/framew55.jpg and masked images are B/w images where the purple area is extracted as white.it is basically a problem where I need to classify purple patches and the rest in masked image! $\endgroup$ – Hima Jan 5 '17 at 11:40
  • $\begingroup$ I don't think you want ML techniques for this task. I think you probably want some form of region detection based on colour values. Try something like this: learnopencv.com/blob-detection-using-opencv-python-c $\endgroup$ – Neil Slater Jan 5 '17 at 13:05
  • $\begingroup$ Even I thought initially but It is said to use ML or AI technique to approach the problem. Moreover why the output files are given along! $\endgroup$ – Hima Jan 5 '17 at 14:09
1
$\begingroup$

I would suggest you use Python with Ski-Image for Image Related operation. For Machine Learning typically for Deep Learning, you should use Keras with theano/Tensorflow backend with GPU capabilities whichever suits you.

In order to generate masked images, you should use data augmentation.Data Augmentation Documentation for Keras

| improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.