bradysm / dataset_splitter

Custom dataset splitter

Image classification Dataset splitter

Python script written to save me time from manually creating a testing set from a single directory filled with classified images. Instead of manually creating a testing set (or even worse using the same training set for testing) I wrote this script in 10 mins to help you create a testing set for your classified images. Simply create dircotry as seen below:

• Root Directory
  • Dataset_dir
    • label_1_images
    • label_2_images
    • ....
    • label_n_images

The script will then take all of the images for each label and create a testing set by splitting the data based on a percentage you define in the script.

Result

• Root Directory
  • Dataset_dir
    • label_1_images
    • label_2_images
    • ....
    • label_n_images
  • testing_set_dir
    • label_1_images
    • label_2_images
    • ....
    • label_n_images

Where a percentage of the labeled images are moved to the testing set which is created for you 😄