Metadata improvements

[kozlov721]

• Added a Category type that can be used with the task metadata to signify the value is categorical as opposed to a general string
• Added new fields to the LuxonisDataset.metadata
  • "categorical_encodings" containing numerical encodings for categorical metadata tasks
  • "metadata_types" containing a mapping from metadata tasks to their python types
  • Added corresponding getters to LuxonisDataset
    • get_categorical_encodings
    • get_metadata_types
• New parameter keep_categorical_as_strings in LuxonisLoader, default is False
  • If False, the loader output will contain the category ids instead of the string values
• Fixed a few small issues with metadata labels

This change allows for better flexibility in luxonis-train for models requiring metadata labels.

Example:

from luxonis_ml.data import Category

def generator():
    yield {
        "task": "vehicles",
        ...,
        "metadata": {
            "color": Category("red")  # one of "red", "green", "blue"
            "license_plate": "xyz",
            "distance": 12.4,
            "id": 4,
        }
    }

dataset = LuxonisDataset(...)
dataset.add(generator)
assert dataset.get_categorical_encodings() == {
    "vehicles/metadata/color": {"red": 0, "green": 1, "blue": 2}}
}
assert dataset.get_metadata_types() == {
    "vehicles/metadata/color": "Category",
    "vehicles/metadata/license_plate": "str",
    "vehicles/metadata/distance": "float",
    "vehicles/metadata/id": "int",
}

[codecov]

Codecov Report

Attention: Patch coverage is 93.93939% with 4 lines in your changes missing coverage. Please review.

Project coverage is 95.77%. Comparing base (a770f26) to head (a49b687).
Report is 1 commits behind head on main.

✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
luxonis_ml/data/datasets/luxonis_dataset.py	91.83%	4 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #234      +/-   ##
==========================================
- Coverage   95.79%   95.77%   -0.03%     
==========================================
  Files          90       90              
  Lines        4785     4828      +43     
==========================================
+ Hits         4584     4624      +40     
- Misses        201      204       +3     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.