2
$\begingroup$

I have a dataset which consists of more than 10000 images. But similar images are grouped together. I mean first 50 images are very alike then the next 50 images are different(not as much similar as the first ones. I am talking about guns specifically) than the first 50 but they are similar between each other. If I choose a batch size of 50, will it lead to worse results? Or It picks random subsets of the dataset to train at a time?

I am new at deep learning if the answer is obvious sorry.

$\endgroup$
1
  • $\begingroup$ For a better answer, please provide those lines of code that configure and consume the batches. $\endgroup$ – Esmailian Mar 13 '19 at 13:59
2
$\begingroup$

Welcome to the site! My understanding is that it does it sequentially. It's usually a good idea to shuffle your dataframe as a pre-processing step and then train on that.

$\endgroup$

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.