Classification Conversion

A guide for bidirectional V1/V2 conversion of image classification annotations.

Data Structure

V1 Structure

# annotations
{
    "id": "cls_1",
    "tool": "classification",
    "isLocked": False,
    "isVisible": True,
    "isValid": True,
    "classification": {
        "class": "cat",
        "confidence": 0.98,
        "breed": "persian"
    },
    "label": ["cat"]
}

# annotationsData (id only)
{
    "id": "cls_1"
}

V2 Structure

{
    "id": "cls_1",
    "classification": "cat",
    "attrs": [
        {"name": "confidence", "value": 0.98},
        {"name": "breed", "value": "persian"}
    ],
    "data": {}  # empty object
}

Conversion Rules

V1 → V2

V1 Field	V2 Field
(none)	data: {} (empty object)
classification.class	classification
classification.{other}	attrs[{name, value}]

V2 → V1

V2 Field	V1 Field
data	(ignored)
classification	classification.class
attrs[{name, value}]	classification.{name: value}

note

The classification tool has no coordinate data, so only id is stored in annotationsData. In V2, data is an empty object {}.

Usage Examples

Basic Conversion

from synapse_sdk.utils.converters.dm import convert_v1_to_v2, convert_v2_to_v1

# V1 classification data
v1_data = {
    "annotations": {
        "image_1": [
            {
                "id": "ClsAbc1234",
                "tool": "classification",
                "classification": {"class": "cat", "confidence": 0.98, "breed": "persian"}
            }
        ]
    },
    "annotationsData": {
        "image_1": [
            {
                "id": "ClsAbc1234"
            }
        ]
    }
}

# Convert to V2
result = convert_v1_to_v2(v1_data)
annotation_data = result["annotation_data"]

# Check V2 result
cls = annotation_data["images"][0]["classification"][0]
print(cls["classification"])  # "cat"
print(cls["data"])  # {}
print(cls["attrs"])  # [{"name": "confidence", "value": 0.98}, {"name": "breed", "value": "persian"}]

Multi-label Processing

# Image with multiple classification tags
v1_multi = {
    "annotations": {
        "image_1": [
            {"id": "cls_animal", "tool": "classification", "classification": {"class": "animal"}},
            {"id": "cls_indoor", "tool": "classification", "classification": {"class": "indoor"}},
            {"id": "cls_pet", "tool": "classification", "classification": {"class": "pet"}}
        ]
    },
    "annotationsData": {
        "image_1": [
            {"id": "cls_animal"},
            {"id": "cls_indoor"},
            {"id": "cls_pet"}
        ]
    }
}

# Convert to V2
result = convert_v1_to_v2(v1_multi)
classifications = result["annotation_data"]["images"][0]["classification"]

for cls in classifications:
    print(cls["classification"])
# animal
# indoor
# pet

Roundtrip Verification

def verify_classification_roundtrip(v1_original):
    """Verify classification roundtrip"""
    # V1 → V2 → V1
    v2_result = convert_v1_to_v2(v1_original)
    v1_restored = convert_v2_to_v1(v2_result)

    # Compare classifications
    orig_cls = v1_original["annotations"]["image_1"][0]["classification"]
    rest_cls = v1_restored["annotations"]["image_1"][0]["classification"]

    assert orig_cls["class"] == rest_cls["class"]

    print("Classification roundtrip verification successful")

verify_classification_roundtrip(v1_data)

Differences from Other Tools

Property	Classification	Bounding Box
Coordinate data	None	Present
V1 annotationsData	id only	coordinate
V2 data	{}	[x, y, w, h]
Purpose	Whole image tagging	Region specification