What is the correct format for training on custom data?

[Callum232310]

Hi, I see in the description the following layout is required for training on a custom dataset:

inputs
└── <dataset name>
    ├── images
    |   ├── 0a7e06.jpg
    │   ├── 0aab0a.jpg
    │   ├── 0b1761.jpg
    │   ├── ...
    |
    └── masks
        ├── 0
        |   ├── 0a7e06.png
        |   ├── 0aab0a.png
        |   ├── 0b1761.png
        |   ├── ...
        |
        └── 1
            ├── 0a7e06.png
            ├── 0aab0a.png
            ├── 0b1761.png
            ├── ...

Let's say 0a7e06.png has one instance of class 0 and no instances of class 1. My question is, is a blank image in the class 1 folder named 0a7e06.png required? Or is 0a7e06.png only required to exist in folder 0?

Thanks