Train, Validation and Test split overlap about labelme2yolo HOT 4 CLOSED

@Arshadoid Thanks for the feedback, you can submit a pr to fix the issue.

from labelme2yolo.

@Arshadoid Thanks for the feedback, you can submit a pr to fix the issue.

@GreatV Thank, will do now. This will be my first contribution :)

from labelme2yolo.

this part of the code is not correct I believe:

train_idxs, test_idxs = train_test_split(
    range(tmp_train_len), test_size=test_size / (1 - val_size))

I think it should be:

train_idxs, test_idxs = train_test_split(
    train_idxs, test_size=test_size / (1 - val_size))

as we want to split already created train indexes further for new train and test. Will test and post the result here.

from labelme2yolo.

@GreatV I just tested and seems like it works correctly now. I commited the updated code to the repo.

1. Verified duplicates with a python script:

• Overlap between Train and Val: 0 files
• Overlap between Train and Test: 0 files
• Overlap between Val and Test: 0 files

2. Amount of combined split files are now matching the total amount.

from labelme2yolo.