1

I'm currently using the cv2.goodFeaturesToTrack() method. However, the corners it returns are somewhat vague and doesn't really do what i wanted wherein it would put some dots on the outline of the character. Here is an attached image of how it worked on my custom dataset: sample image

corners = cv2.goodFeaturesToTrack(crop, 8, 0.02, 10)

If cv2.goodFeaturesToTrack() is not a good way of extracting features from handwritten characters, what are some of the better ways to do it? Thank you

superfab
  • 78
  • 8
  • `cv2.goodFeaturesToTrack` is to find good corners, not suitable for characters. Hog feature is a good choice. – Kinght 金 Nov 19 '17 at 14:15

1 Answers1

1

Use Neural Networks you can preload trained models also in open cv. Here is short intro into the topic. Its a 1 to 1 tutorial :) Hope you like it. https://www.pyimagesearch.com/2014/09/22/getting-started-deep-learning-python/

Max Krappmann
  • 490
  • 5
  • 19