DM Schema V1/V2 Converter

A bidirectional converter between DM Schema V1 and V2 formats.

Overview

The DM Schema Converter transforms annotation data between V1 and V2 formats:

• V1 → V2: Convert legacy V1 data to modern V2 structure
• V2 → V1: Reverse convert V2 data to V1 format (legacy system compatibility)

Key Features

Separated Output Structure

V1→V2 conversion returns two separate parts:

• annotation_data: V2 common annotation structure (id, classification, attrs, data)
• annotation_meta: Preserved V1 top-level structure (meta information)

This separation enables:

• Use annotation_data alone for V2 common format
• Combine with annotation_meta for complete V1 restoration

Supported Tools

Tool	V1 → V2	V2 → V1	Media Type	Notes
bounding_box	✅	✅	image	rotation preserved
polygon	✅	✅	image	point IDs auto-generated
polyline	✅	✅	image	point IDs auto-generated
keypoint	✅	✅	image	single point coordinate
3d_bounding_box	✅	✅	pcd	PSR coordinate system
segmentation	✅	✅	image/video	image: pixel_indices, video: section
3d_segmentation	✅	✅	pcd	point indices
named_entity	✅	✅	text	NER tagging
classification	✅	✅	image	no data (empty object)
relation	✅	✅	image/text	annotation relationships
prompt	✅	✅	prompt	prompt input
answer	✅	✅	prompt	answer output

Quick Start

Installation

pip install synapse-sdk

V1 → V2 Conversion

from synapse_sdk.utils.converters.dm import convert_v1_to_v2

# V1 data
v1_data = {
    "annotations": {
        "image_1": [
            {
                "id": "ann_1",
                "tool": "bounding_box",
                "classification": {"class": "person"}
            }
        ]
    },
    "annotationsData": {
        "image_1": [
            {
                "id": "ann_1",
                "coordinate": {"x": 100, "y": 200, "width": 150, "height": 100}
            }
        ]
    }
}

# Convert to V2 (separated result)
result = convert_v1_to_v2(v1_data)

annotation_data = result["annotation_data"]  # V2 common structure
annotation_meta = result["annotation_meta"]  # V1 top-level structure

V2 → V1 Conversion

from synapse_sdk.utils.converters.dm import convert_v2_to_v1

# Complete conversion (annotation_data + annotation_meta)
v1_restored = convert_v2_to_v1(result)

# Convert with annotation_data only (uses defaults)
v1_basic = convert_v2_to_v1({"annotation_data": annotation_data})

Detailed Guides

Image Tools

• Bounding Box Conversion - Bounding box annotation conversion
• Polygon Conversion - Polygon annotation conversion
• Polyline Conversion - Polyline annotation conversion
• Keypoint Conversion - Keypoint annotation conversion
• Segmentation Conversion (Image) - Image segmentation conversion
• Classification Conversion - Classification annotation conversion
• Relation Conversion - Annotation relationship conversion

Video Tools

• Segmentation Conversion (Video) - Video segmentation conversion

PCD (Point Cloud) Tools

• 3D Bounding Box Conversion - 3D bounding box conversion
• 3D Segmentation Conversion - 3D segmentation conversion

Text Tools

• Named Entity Recognition Conversion - NER annotation conversion

Prompt Tools

• Prompt Conversion - Prompt annotation conversion
• Answer Conversion - Answer annotation conversion

Developer Documentation

• Developer Guide - How to add new tools

API Reference

convert_v1_to_v2(v1_data)

Converts V1 data to V2 format.

Parameters:

• v1_data: DM Schema V1 format data

Returns:

• V2ConversionResult: Dictionary containing annotation_data and annotation_meta

convert_v2_to_v1(v2_data, annotation_meta=None)

Converts V2 data to V1 format.

Parameters:

• v2_data: DM Schema V2 format data or V2ConversionResult
• annotation_meta: Optional V1 top-level structure passed separately

Returns:

• DM Schema V1 format data