Tasks

Multiclass classification

Multiclass classification models predict the single best fitting category for an input image among a predefined set of possibilities.
The set of possible categories can be ordered (e.g. ["bad", "ok", "good"]), in which case the task could be called "ordinal classification".

Dataset format

Datasets follow this structure:

endpoint_url/bucket
├── prefix/images/
├── prefix/annotations.yaml
└── prefix/metadata.yaml

Dataset images are placed directly inside images/ (subdirectories are ignored).
The metadata file looks something like this:

metadata.yaml
task: multiclass classification
annotations: annotations.yaml
categories: [cat1, cat2, cat3]

The annotations field specifies the name of the file containing the ground truth annotations.
Here's an example of annotations file:

annotations.yaml
000.jpg: cat2
001.jpg: cat3
002.jpg: cat1
# ...
Only the categories specified in the metadata should be used.
All images must be assigned a category (or null, in which case they would be ignored).