Import Labels
Overview
The Import Labels Plugin allows you to import pre-labels into a project from a CSV file. This plugin is designed to support bulk ingestion of labels generated by external tools, models, or annotation pipelines and map them correctly to existing assets in your project.
Import Labels
Plugin Functionality
The plugin reads a CSV file where each row represents a single asset and imports labels based on the configured parameters. Depending on the configuration, it can:
Match CSV rows to existing project assets using an external ID column
Import classification labels from one or more CSV columns
Import segmentation/polygon annotations from mask files referenced in the CSV
Map mask values to label classes using project colors or custom mappings
Overwrite existing labels or append new ones
Process imports in customized batches for better performance.
Example Configurations
Import Classification Annotations
Import Segmentation Annotations
Supported Data Types
Compatible with all data types available in AngoHub.
Supported Annotation Tools
Classifications
Radio
Checkbox
Single-Select Dropdown
Multi-Select Dropdown
Text
Tools
Segmentation
Polygon
Following tools are not supported by the plugin:
Nested classifications, multiple classifications, frame specific classifications
Frame-specific segmentation and polygon tools
Nested classifications within tools
Plugin Configuration
You may vary a number of settings related to your export from the Config JSON field. Each option is detailed below:
"external_id_column": Name of the CSV column that contains external IDs used to match imported labels to existing assets in the project.
Example:
"external_id_column": "External ID"
"classification_columns": List of CSV column names that contain classification values to import. Each column is mapped to a classification schema.
Example:
"classification_columns": ["Color", "Category"]
"classification_schema_ids": List of classification schema IDs that define how values from classification_columns are interpreted and mapped during import. The order must match classification_columns.
Example:
"classification_schema_ids": ["12345", "12346"]
"mask_column": Name of the CSV column that contains segmentation mask references. The value can be a file path or URL pointing to the mask file.
Example:
"mask_column": "Mask"
"mask_color_mapping": Defines how colors in segmentation masks are mapped to label classes. Use project colors or provide a custom mapping.
Options:
"mask_color_mapping": "project_colors"
"mask_color_mapping": {"Car" : [244, 67, 54], "Person" : [3, 169, 244]}
"mask_color_mapping": {"Car" : [1], "Person" : [2]}
"segmentation_mode": Controls how segmentation annotations are created from masks. "instance_segmentation" creates a separate object for each connected region per class and "semantic_segmentation" creates a single object for each class.
Options:
"segmentation_mode": "instance_segmentation"
"segmentation_mode": "semantic_segmentation"
"polygon_simplification_tolerance": Tolerance used when converting masks to polygons. Higher values reduce the number of polygon points at the cost of geometric precision.
Example:
"polygon_simplification_tolerance": 0.1
"storage_id": Identifier of the storage location where mask files referenced in the CSV are stored. Used to resolve file paths or storage keys during import.
Example:
"storage_id": "1234"
"overwrite": If true, existing labels for matching assets will be replaced by the imported labels. If false, imported labels are added while preserving existing labels.
Example:
"overwrite": true
"overwrite": false
"ignored_schema_ids": List of schema IDs to ignore during import. Any tools or classifications associated with these schemas will be skipped, even if present in the CSV.
Example:
"ignored_schema_ids": ["1234", "1235"]
"upload_batch_size": Number of rows from the CSV to process per batch. Larger batch sizes improve throughput but increase memory usage and may hit API limits.
Example:
"upload_batch_size": 100
Last updated