2
$\begingroup$

I'm working on a machine learning project, Images classification (shape: 100 x 100)-> (vector of 10000), I did some pre-processing before applying decision trees algorithm , I got an accuracy of 55 % I tried to change parameters but the accuracy did not increase, Can someone suggest something to do in pre-processing exept what I did :

  • Denoising Images: Images are extremely noised, so I only kept relevant information in each image.(it works well)
  • Centring Images: I translated the relevant values for all images to the top left to keep relevent information in same part of my vectors
  • Resizing Images: I resized each image to (40,40) -> (1600 vector) to make the inputs smaller

Thank you

$\endgroup$

2 Answers 2

4
$\begingroup$

Why are you convinced that the issue is in pre-processing? It could be in many other spots or a combination of steps in your process.

If I were you, I would revisit WHY you're using a decision tree. There are many other approaches to image processing and neural networks are far superior - in most cases - to decision trees.

$\endgroup$
2
$\begingroup$

If you are focusing just on pre-processing, you should try different techniques of image augmentation, which include standardization, rotation, flipping, shifts and shears etc. Here are 2 good links:

https://machinelearningmastery.com/image-augmentation-deep-learning-keras/

https://medium.com/nanonets/how-to-use-deep-learning-when-you-have-limited-data-part-2-data-augmentation-c26971dc8ced

These techniques can significantly enhance accuracy.

Moreover, you may be losing some information converting from 100x100 to 40x40. Try with original size and you may get better accuracy (although it will be more intensive computationally).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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