Contents Menu Expand Light mode Dark mode Auto light/dark mode
3LC Logo
Blog
Company
Documentation
Discord
  • Home
  • Introduction
  • Quickstart
  • User Guide
    • 3LC Dashboard
      • Example
      • Panels
      • Pages
      • Views
      • Operations
      • Keyboard Shortcuts
      • GPU Acceleration
    • tlc Python Package
      • Core Concepts
        • Objects, URLs, and Schemas
        • Tables
        • Runs
      • Register Datasets
      • Collect Metrics
      • 3LC Project Structure
      • SampleType and structure
      • Sample Weights
      • URL Details
      • Embeddings
      • Combining Data and Models in 3LC
    • 3lc Command Line Interface
    • 3LC Object Service
      • Deployment Guide
      • Indexing
    • Integrations
      • PyTorch
      • Hugging Face
      • Detectron2
      • YOLOv5
      • YOLO
      • PyTorch Lightning
      • Google Colab
      • AWS SageMaker Studio
    • Configuration
    • Sharing 3LC Tables and Runs
    • 3LC File Locations
    • 3LC Environment Variables
    • System Requirements
    • Glossary
    • 3LC Account
      • 3LC Account and Workspaces
      • API Key
    • 3LC Enterprise Customer Managed
      • Quickstart
      • 3LC Customer Managed Architecture
      • Licensing
  • Example Notebooks
    • MNIST
    • FashionMNIST
    • CIFAR-10
    • Detectron2 Balloons
    • Detectron2 COCO128
    • Hugging Face IMDb
    • Hugging Face fine-tuning
    • Hugging Face CIFAR-100 Embeddings
    • PyTorch Lightning SegFormer
    • Bounding Box Notebooks
      • Per Bounding Box Metrics
  • Tutorials
    • YOLO Hardhat Detection
    • GitHub Tutorial Notebooks
  • How-To
    • Basics
      • Get started with 3LC Python package
      • Create a Table
      • Use aliases when making a Table
      • Export a Table
      • Create a Run
      • Open a Project/Table/Run
      • View Run's details
      • Record and view hyperparameters
      • Manipulate metrics/example tables
      • Create a chart
      • Tweak a chart
      • Overlay properties on a chart
      • Use polygon selections
      • Use filters
      • Show partially filtered-in samples across epochs
      • Create/edit an editable column
      • Edit a label column
      • Commit edits
      • View images in grid view
      • Create a subset table
      • Create a reduced table
      • Create a virtual column
      • Create a confusion matrix
      • Open two Runs in a session
      • Modify global settings
      • Find keyboard shortcuts
      • Use workflows
      • Use the N-dimensional travel distance workflow
    • Bounding Boxes and Masks
      • Create a BB/mask and image overlay chart
      • Navigate BBs
      • Set BB display properties
      • Derive BB-related virtual properties
      • Use BB-based filters
      • Create a new class
      • Create and edit BBs
      • Convert predicted BBs to ground truth
      • Use NMS
      • Edit a semantic segmentation mask
      • Accept a semantic segmentation mask
      • Edit instance segmentation masks
      • Convert predicted instance segmentation masks to ground truth
  • Frequently Asked Questions
  • Python API
    • tlc
      • client
        • bulk_data_url_utils
        • data_format
        • helpers
        • log
        • operations
          • table_operations
        • reduce
          • pacmap
          • reduce
          • reduction_method
          • umap
        • sample_type
        • torch
          • metrics
            • collect
            • metrics_aggregators
              • aggregator
              • mean_aggregator
            • metrics_collectors
              • bounding_box_metrics_collector
              • classification_metrics_collector
              • embeddings_metrics_collector
              • functional_metrics_collector
              • metrics_collector_base
              • segmentation_metrics_collector
            • predictor
      • core
        • builtins
          • constants
            • bool_roles
            • column_names
            • composite_roles
            • defaults
            • display_importances
            • number_roles
            • paths
            • reshape_strategies
            • schema_fields
            • string_roles
            • units
            • values
            • widgets
          • schemas
            • bounding_box_list_schema
            • rgb_image_schema
            • row_count_schema
            • schemas
            • top_n_prediction_schema
            • voxelset_schema
          • types
            • bb_crop_interface
            • bounding_box
        • exceptions
        • export
          • exporter
          • exporters
            • coco
            • csv
            • default_json
        • init_global_objects
        • json_helper
        • lru_cache
        • object
        • object_reference
        • object_registry
        • object_type_registry
        • objects
          • mutable_object
          • mutable_objects
            • configuration
            • run
          • null_object
          • table
          • tables
            • from_python_object
              • table_from_pandas
              • table_from_pydict
              • table_from_torch_dataset
            • from_table
              • dimensional_reduction_table
              • edited_table
              • filtered_table
              • filtered_table_criteria
                • all_filter_criterion
                • any_filter_criterion
                • bool_filter_criterion
                • filter_criterion
                • filter_criterion_factory
                • freetext_filter_criterion
                • integer_set_filter_criterion
                • logical_not_filter_criterion
                • numeric_range_filter_criterion
                • region_filter_criterion
                • text_filter_criterion
              • pacmap_table
              • reduced_table
              • rolled_up_table
              • schema_helper
              • subset_table
              • umap_table
              • unrolled_table
            • from_tables
              • joined_table
            • from_url
              • table_from_coco
              • table_from_csv
              • table_from_parquet
              • table_from_row_cache
              • table_from_yolo
              • utils
            • generators
              • random_table
            • in_memory_columns_table
            • in_memory_rows_table
            • null_overlay
            • null_table
            • system_tables
              • config_scan_iterator
              • file_scan_iterator
              • indexing
              • indexing_table
              • indexing_tables
                • config_indexing_table
                • lineage_helper
                • run_indexing_table
                • table_indexing_table
                • table_info
              • log_table
              • project_scan_iterator
              • single_dir_scan_iterator
              • timestamp_helper
        • operations
          • bb_operations
          • operation
        • schema
        • schema_helper
        • serialization_version_helper
        • table_row_serializer
        • table_row_serializer_registry
        • table_row_serializers
          • parquet_table_row_serializer
        • type_helper
        • url
        • url_adapter
        • url_adapter_registry
        • url_adapters
          • abfs_url_adapter
          • api_url_adapter
          • file_url_adapter
          • file_url_adapter_win
          • fsspec_url_adapter
          • gcs_url_adapter
          • http_url_adapter
          • s3_url_adapter
        • writers
          • metrics_writer
          • table_writer
      • integration
        • detectron2
          • detectron_metrics_collection_hook
          • metrics_collection_hook
          • register_coco_instances
          • umap_reduce_embeddings_hook
        • hugging_face
          • features
          • table_from_hugging_face
          • trainer
          • trainer_deprecated
        • pytorch_lightning
          • module_decorator
    • tlcconfig
      • constants
      • default_locations
      • logger_configurator
      • option_loader
      • options
  • Release Notes
    • 3LC Python Package Release Notes
      • 3LC Python Package Version 2.13
      • 3LC Python Package Version 2.12
      • 3LC Python Package Version 2.11
      • 3LC Python Package Version 2.10
      • 3LC Python Package Version 2.9
      • 3LC Python Package Version 2.8
      • 3LC Python Package Version 2.7
      • 3LC Python Package Version 2.6
      • 3LC Python Package Version 2.5
    • 3LC Dashboard Release Notes
      • 3LC Dashboard Version 2.6
      • 3LC Dashboard Version 2.5
    • Earlier Versions
      • 3LC Version 2.3 Release Notes
      • 3LC Version 2.2 Release Notes
      • 3LC Version 2.1 Release Notes
      • 3LC Version 2.0 Release Notes
      • 3LC Version 1.0 Release Notes
Back to top

How to derive BB-related virtual properties¶

BB-related virtual properties are essentially a group of virtual columns derived from either BBS or BBSpredicted or both. These virtual properties can be used the same way as the collected BB metrics. All available virtual properties are listed under Derive virtual column in the popup menu when RightClick’ing a selected bounding box column.

Some commonly used virtual properties of BBS or BBSpredicted include:

  • Rectangles(absolute): An array of four absolute values in form of [min_x, min_y, max_x, max_y] for each BB

  • Rectangles(relative): An array of four relative values in form of [min_x, min_y, max_x, max_y] for each BB

    • Area: Area for each BB

    • Aspect: Aspect ratio for each BB

    • Height: Height of each BB

    • Width: Width of each BB

  • Overlap ratio: Sum of overlapped areas of a given BB with all other BBs divided by area of this BB; value can be greater than 1

  • Unique overlap ratio: Sum of unique overlapped areas of a given BB with all other BBs divided by area of this BB; value is equal to or less than 1

Some commonly used virtual properties derived from selecting both BBS and BBSpredicted include:

  • IoU: Intersection over union between ground truth and predicted BBs. This is an order-dependent virtual column, which means that it will be computed on the basis of and assigned to the first selected column.

  • FN: False negatives with default 0.5 of IoU threshold, which is adjustable; FNs are assigned to ground truth BBs regardless of the selection order of the two columns.

  • FP: False positives with default 0.5 of IoU threshold, which is adjustable; FPs are assigned to predicted BBs regardless of the selection order of the two columns.

  • TP: True positives with default 0.5 of IoU threshold, which is adjustable; TPs are assigned to predicted BBs regardless of the selection order of the two columns.

    • Sum: a chained property under FN/FP/TP; it can be used to calculate the total FNs/FPs/TPs for each sample.

To adjust the IoU threshold for FN/FP/TP, select the derived FN/FP/TP column, RightClick and hover over Tweak virtual column, and then adjust the value either by typing in the box or dragging the slider.

FN/FP/TP/IoU: virtual properties vs collected metrics

Like all virtual columns created in the Dashboard, FN/FP/TP/IoU virtual properties are dynamic and re-computed e.g. as ground truth BBs are edited or IoU thresholds are adjusted. On the other hand, you may have collected FN/FP/TP/IoU metrics during metrics collection that you are also able to display in the Dashboard. Keep in mind that such metrics are static (e.g. they will have the fixed IoU threshold used when collected) and they cannot be changed in the Dashboard. The dynamic character of the Dashboard virtual properties makes their use especially versatile and flexible.

Next
How to use BB-based filters
Previous
How to set BB display properties
Copyright © 2025, The 3LC Team

We use cookies to improve 3LC’s site. Some cookies are necessary for our website and services to function properly. Other cookies are optional and help personalize your experience, including advertising and analytics. You can consent to all cookies, decline all optional cookies, or manage optional cookies. Without a selection, our default cookie settings will apply. You can change your preferences at any time. To learn more, check out our Cookie Policy.