Single file output format

[hlky]

Currently output is saved to individual files.

Saving all outputs to a single file will be especially beneficial for large scale captioning.

There are a few ways this could work:

• dictionary with filenames as keys and each contains a dictionary with tasks as keys
• dictionary with tasks as keys and each contains a dictionary with filenames as keys
• single task per file, dictionary with filenames as keys, with or without task

The use of dictionaries is preferred to enable updates/appending. We can also track which files have already been captioned.

Examples

Option 1

{
"filename.jpg": {
  "<CAPTION>": "caption text"
  }
}

Option 2

{
"<CAPTION>": {
  "filename.jpg": "caption text"
  }
}

Option 3

{
  "filename.jpg": "caption text"
}

{
"filename.jpg": {
    "<CAPTION>": "caption text"
  }
}