Datasets:
Wrong(?) images information inside "images"
I started using these annotations for a university project, and I noticed that in the "images" information within the JSON files there are also data from the previous manga in alphabetical order.
For example, inside AisazuNihaIrarenai.json, the first image entries refer to ARMS, which is the previous manga in alphabetical order. This means that YumeNoKayoiji.json, which is the last one alphabetically, contains all the image data.
This appears to be some kind of mistake introduced by an automatic tool, but I wonder if I am missing something.
Moreover, I noticed that some "images" entries contain one more image than the Manga109 dataset I have access to. For example, ARMS.json contains 81 entries in "images", while there are only 80 images in the dataset. Additionally, there are no annotations for the 81st image, so it appears to be an extra entry.
I wonder if there are different versions of the datasets, or if there is another explanation.