kognic-openlabel

Kognic OpenLABEL

Installation

Install the Kognic OpenLABEL package from pip with

pip install kognic-openlabel

Serialization and deserialization

Since all models inherit from pydantic's BaseModel, serialization and deserialization from dicts or json strings are relatively straight forward.

data = {
    "openlabel": {
        "metadata": {
            "schema_version": "1.0.0"
        }
    }
}

import kognic.openlabel.models as OLM

# Deserialize dict
openlabel_annotation = OLM.OpenLabelAnnotation.model_validate(data)

# Serialize to json
json_data = openlabel_annotation.model_dump_json(exclude_none=True)

# Deserialize json
openlabel_annotation = OLM.OpenLabelAnnotation.model_validate_json(json_data)

# Serialize to dict
dict_data = openlabel_annotation.model_dump(exclude_none=True)

Further reading

https://www.asam.net/project-detail/asam-openlabel-v100/

Changelog

[2.0.2] - 2024-07-26

• Fix so that model_dump works correctly with ontologies

[2.0.1] - 2023-11-07

• Use model_rebuild instead of deprecated update_forward_refs

[2.0.0] - 2023-11-06

• Bump pydantic to >2

[1.0.1] - 2023-07-05

• Pin pydantic version to <2

[1.0.0] - 2022-10-18

• Annotell becomes Kognic

[0.1.4] - 2022-01-24

• Improved serializability for enum classes

[0.1.3] - 2022-01-04

• Fixed issues with version 0.1.2

[0.1.2] - 2021-12-29

• Updated several fields with multiple types to fix issues with serialization and deserialization. For example, the coordinates on Poly2d objects were previously always parsed to strings. This update means that an attempt to parse them to floats is made. If this fails the they will be parsed to strings.

[0.1.1] - 2021-11-24

• Updated stream properties model generation to be nicer to work with

[0.1.0] - 2021-11-18

• Updated json schema and model to be true to the 1.0.0 release of openlabel. Previously it was based on the release-candidate