Hello, I am working on document images and i have got the dataset also. Now i wish to extract the features from those images by using any machine learning techniques. as you have mentioned in your project feature extraction and machine learning. Can you help me in this context.
Thank you,
With regards,
SATISH KUMAR
You can use conventional network on your image data set then the the output on these kind networks gives you the feature maps.
thanks, Amir for your valuable suggestion. can you elaborate it
Hi Satish,
The conventional network is an advanced classification method. However, before applying such method, you need to extract the image features, such as extracting MPEG7 color layout features, extracting a fuzzy 64-bin histogram, etc.
HTH.
Samer
Thanks for the suggestion Sameer, i got the grayscale images of sized 128x28,256x256,512x512 and 1024x1024 each size got 6000 images. for these grayscale images i need to extract features, can you suggest which is best method to extract the features by using CNN.
No worries, Satish.
Blindly guessing best algorithm without knowing the data is not a good idea. Therefore, I recommend trying several methods to extract data features and select the one produces with CNN a reasonable result. For now, I can only suggest using methods that can:
- extract MPEG7 edge histogram features method, and/or
- extract PHOG features, and/or
- extract a pyramid of rotation-invariant local binary pattern histograms
Of course, there are many other methods that you could try in addition to the ones I mentioned.
HTH.
Samer
Hello,
Kindly go through the paper attached.
For feature extraction , there are many techniques, however, i had used 3-level DWT and that gives a satifactory result.
Regards,
Rebika Rai
- Badges
- Science topic
- Similar topics
- Computer Science
- Data Mining