ObjectDetectionAnnotation

Represents a single COCO-style object-detection or instance-segmentation annotation for one object. Required fields are id, image_id, and category_id; optional fields include segmentation, bbox, area, iscrowd, and score. This datatype is used as the schema for each element in ObjectDetectionAnnotations within the Telekinesis ecosystem.

Field

Required

Field	Type	Description
id	int	Annotation identifier (int).
image_id	int	Image identifier (int).
category_id	int	Category identifier (int).

Optional

Field	Type	Description
segmentation	list[list[float]], list[np.ndarray], or dict	Polygons (list of polygons) or RLE dict with counts and size. Default [].
bbox	list[int] or np.ndarray	Bounding box [x, y, width, height]; length 4. Default [0, 0, 0, 0].
area	float	Area of the segment. Default 0.0.
iscrowd	int	Whether the annotation is a crowd region. Default 0.
score	float or None	Detection or confidence score. Default None.

Methods

Method	Description
to_dict()	Returns a dictionary with keys: id, image_id, category_id, segmentation, bbox, area, iscrowd, score.

Example

From the datatypes examples:

from datatypes import datatypes
from loguru import logger

# ------------------------------------------------
# 1. Create ObjectDetectionAnnotation instance
# ------------------------------------------------
ann = datatypes.ObjectDetectionAnnotation(
    id=1,
    image_id=0,
    category_id=1,
    segmentation=[[10.0, 10.0, 50.0, 10.0, 50.0, 50.0, 10.0, 50.0]],
    bbox=[10, 10, 40, 40],
    area=1600.0,
    iscrowd=0,
)

# ------------------------------------------------
# 2. Access data via to_dict
# ------------------------------------------------
d = ann.to_dict()
logger.info("ObjectDetectionAnnotation id={}", d["id"])
logger.info("ObjectDetectionAnnotation image_id={}", d["image_id"])
logger.info("ObjectDetectionAnnotation category_id={}", d["category_id"])
logger.info("ObjectDetectionAnnotation bbox={}", d["bbox"])
logger.info("ObjectDetectionAnnotation area={}", d["area"])
logger.info("ObjectDetectionAnnotation score={}", d.get("score"))