2025.1.2

• Improved visual alignment of the toe and foot segments in the skeleton visualization.

  • This update enhances display accuracy only, no model changes

2025.1.1

• Added support for NVIDIA GeForce RTX 50 series graphics cards

2025.1.0

Theia3D

• Theia theme and rebranding

• Improved error messages in dialogs, including links to troubleshooting pages

• Added links to web pages with additional information for tools and interfaces throughout

• Organize Videos Advanced tab:

  • Added ability to edit the subject and actions directly and export a csv file

• Added Help > Submit Support Ticket tool to allow easy and direct support ticket submissions

Theia3D Batch

• Theia theme and rebranding

• Save Videos option in Batch processing

  • Select display options and specify camera views to save

• Display Errors table added to simplify viewing warnings and errors

• File > Move C3D files

  • Output C3D files can be copied from the root folders to within the parent C3D folder after the batch process has been completed

• Added Help > Submit Support Ticket tool to allow easy and direct support ticket submissions

What Will You Do?

Theia3D is the premier solution for accurate, reliable, and generalizable markerless motion capture. With our software at your fingertips, you'll unlock the power to analyze human movement like never before. Whether you're a researcher exploring human movement, a sports biomechanist tasked with unleashing athletes' full potential, or a product developer working on the next generation of athletic equipment, Theia3D empowers you to:

• Obtain motion data without the need for suits, sensors, or markers.

• Transform raw video footage into precise, 3D models of human movement.

• Integrate effortlessly with leading motion capture software.

• Dive into the future of motion capture and elevate your project to new heights.

Using this Documentation

The reference guides contain detailed information about the interface, settings, and tools available in Theia3D. They are excellent resources for advanced users looking to fine-tune workflows and optimize Theia3D for their specific use cases.

Contact Us

Overview

Theia3D is an advanced deep learning algorithm-based software which runs and processes data on your local PC. This means that there are specific requirements in order for it to operate properly and perform as intended, which are laid out below.

GPU

One or more CUDA-capable NVIDIA graphics card(s) with the following:

• At least 8 GB of memory

CPU

RAM

Installation Steps

After installing Theia software, it is necessary to activate the license prior to using the software. To activate your Theia3D license, run Theia3D as an administrator by right-clicking on the icon and choosing Run as administrator.

If the PC is connected to the internet, the license will automatically activate and Theia3D will open.

After contacting Theia Support, enter the Activation Code as provided by our team and click Activate. The license activation will then be completed, and Theia3D will open.

Theia3D Basics

Theia3D is a markerless motion capture solution that utilizes synchronized video data to produce accurate and reliable 3D pose estimates of humans visible within the video data. It leverages deep learning algorithms trained to identify humans and accurately predict the 2D positions of 124 keypoints on the human body, in every video frame of every camera. By fitting a scaled subject-specific inverse kinematic model to the keypoint predictions, the human’s pose is reconstructed in 3D and tracked throughout their movement. This data-driven approach results in a robust solution that is generalizable across environments and movements, allowing the accessible collection of high quality 3D motion capture data where it was previously impossible.

Here, we describe the basic framework for collecting markerless motion capture data using Theia3D; for more detailed instructions, please refer to the appropriate section of this documentation and any accompanying videos.

Calibration

The calibration methods supported by Theia3D include:

• Chessboard calibration, using a large printed chessboard pattern to automatically obtain intrinsic and extrinsic camera parameters.

• Third-party calibrations, such as wand calibrations implemented by third-party motion capture hardware suppliers.

• Object calibration, using an object with known dimensions or with precisely measured key point positions to manually obtain extrinsic camera parameters.

Movement Data Collection

‍Data Processing

Or, navigate to the relevant section of this documentation and you'll find any specific videos embedded there.

Intrinsic lens calibration trials are used to determine parameters associated with the camera lenses and are used to correct for distortion and other visual effects. Lens calibration trials are required for all camera systems that make use of a chessboard calibration method, except for Sony RX0-ii cameras. Users with wand calibration-capable systems (e.g. Qualisys Miqus, Vicon FLIR Blackfly S systems) are not required to complete lens calibration unless they are planning to use the chessboard calibration method.

Recommendations:

• Record lens calibration trials at a relatively low frame rate (20-30 Hz) to reduce file size and processing time.

• Keep the chessboard as flat as possible during the calibration trial.

• Use a computer monitor facing the person performing the calibration to provide visual feedback during the calibration.

• Move slowly and deliberately to prevent chessboard blur.

• Lens calibrations can be performed for smaller groups or individual cameras and merged afterwards. If you are finding it challenging to calibrate all at once, try recording separate groups.

Recording Lens Calibrations:

1. Place all cameras side by side on a desk or table, facing the same direction and capturing as similar views as possible. It is generally easier to calibrate the lenses with the cameras in 'landscape' orientation.

2. Stand at a distance where the chessboard occupies approximately 1/4 of the camera views.

3. Begin the recording.

4. Slowly move the chessboard in a systematic grid pattern, covering the entire field of view for every camera. Ensure the chessboard goes slightly beyond every edge of every camera field of view, and every corner.

1. Take a step back, and repeat Step 4., covering the camera fields of view again.

2. While covering the field of view, angle the chessboard slightly in multiple directions, varying its orientation throughout.

3. When you are confident that the entire field of view has been covered for all cameras, end the recording.

GPU Selection‍

Select which graphics card(s) to use when running Theia3D.

Person Detector Size

Select which person detector to use when running Theia3D, and set the Skip frames count.

• Small: fastest, may be slightly less accurate than Default or Large.

• Default: recommended.

• Large: slowest, may be slightly more accurate than Default or Small (currently disabled).

Skip frames sets the frame interval used during the analysis process to identify and track people. Default is 1, Maximum is 5. Increasing the Skip Frames value will speed up the analysis. The Skip frames value should not be used for fast movements recorded at relatively low frame rates, or movements with multiple closely-interacting individuals, as the person identification may be affected.

Joint Detector Size

Select which joint detector to use when running Theia3D.

• Small: fastest, may be slightly less accurate than Large (currently disabled).

• Large: (default) slowest, may be slightly more accurate than Small.

• Add Hands: (available as a license add-on by special request) Include hand model, which improves hand tracking through localization. Requires camera views that are positioned specifically to capture the hands and upper body. Please reach out to support@theiamarkerless.com for more information.

Listen Mode

Available as a license add-on by special request.

Listen Mode allows a directory to be selected where data is saved during data collection. The selected directory will be actively monitored for new, unprocessed data which will be automatically analyzed as it is detected. This allows data to be processed during an active data collection session, enabling all processing to be completed by the end of the collection session.

Camera Setup

Camera system setups differ from location to location, and may be subject to challenging data collection environment constraints. General recommendations for setting up your camera system include:

• Cameras as close as possible to the capture volume, while ensuring the entire capture volume remains within view for all cameras.

• Cameras 4-8 feet (1-2.5 m) above the ground.

• Avoid partial views of subjects, such as lower body or upper body only.

• Avoid unusual camera views, especially those positioned very high or very low, or extremely tilted.

• Aim for a symmetrical camera setup that surrounds the entire capture volume, such as a circle, oval, or rectangle.

Camera Settings

The camera settings you choose will depend on the movements being collected. Selection of appropriate camera settings is crucial for collecting clear, crisp video images. The most important camera settings to consider include:

• Shutter speed / Exposure: should be set to ensure the subject and their body segments are not blurry during the movement. Faster shutter speeds or shorter exposures will capture images with less movement blur, but may reduce the image brightness.

Subject Attire

Theia3D provides 3D pose estimates that are robust to changes in attire. General recommendations for subject attire include:

• Body-fitting clothing. Each limb should be discernible from the rest of the body.

• Clothing should provide rich visual features, such as visible creasing, shadows, or other textures.

• Lighting is often more important than attire color, as it impacts the visual richness of the attire. Under adequate lighting conditions black attire is acceptable, but lighter colors generally provide more visual features, especially in low-light environments.

Camera Setup & Position

When setting up cameras, use full circle or oval configurations with multiple sagittal and frontal views. Position cameras at a height of 4 to 8 feet (1 to 2.5 meters).

Aiming for the participant to appear approximately 500 pixels tall in the frame.

• ~½ frame height (landscape @ 1080p)

• ~¼ frame height (portrait @ 1080p)

Avoid unusual angles and strive to keep all cameras flat and level for consistent data capture.

Lighting

​Proper lighting is crucial for high-quality video data in markerless motion capture. Outdoors, be mindful of the sun's position to avoid harsh backlighting and adjust camera settings to prevent overexposed or dark images. Indoors, ensure adequate lighting, as ambient levels are typically lower; aim for at least 500 lux, with 1,000 lux preferred. Be cautious of natural light sources like windows, which can cause silhouette effects.

Use light sources with a color temperature between 3,500K and 6,500K and a Color Rendering Index (CRI) of 70 or higher for accurate color representation. Diffuse, non-directional lighting minimizes shadows and glare, enhancing video quality. Avoid light sources that flicker near your recording frame rate to prevent inconsistent brightness

Frame Rate

​Selecting the appropriate frame rate is crucial for accurately capturing motion in data collection. A frame rate too low may miss subtle movements, while an excessively high frame rate can degrade video quality and result in unnecessarily large data files and long processing times. To determine the optimal frame rate, start with a literature review to understand standard practices for similar studies. Next, assess your camera hardware's capabilities, ensuring it can support the desired frame rate without compromising resolution. Some cameras offer 'windowing' features to increase frame rates by capturing only a portion of the sensor; however, this should be used cautiously to avoid missing parts of the subject. Conduct pilot studies to test different frame rates, reviewing the clarity and quality of the captured movements. Be mindful that higher frame rates may require more lighting and can lead to issues like frame drops due to increased data throughput. Balancing these factors through thorough testing and consideration will help you select a frame rate that meets your experiment's needs.

Extrinsic object calibration trials are used to determine the position and orientation of every camera in your system relative to the desired global coordinate system. Extrinsic object calibration uses the static position of a calibration object with known dimensions or known positions of specific key points on the object. These 3D dimensions or positions must be measured with high precision. The calibration object should be sufficiently large or the calibration key points should be spaced far apart within the capture volume and most points should be visible in every camera view. The key points can be coplanar or can vary in all three global dimensions.

A calibration object file must be created for your calibration object, containing 3D coordinates of the object key points. Each line of the file must contain three comma-separated values representing the global x-, y-, and z-coordinates of the key point, in millimeters. Each key point should be on its own line.

Object calibrations must be performed at the resolution that will be used to record your movement data. If you will be using more than one resolution during your data collection, you must record object calibration trials at each resolution.

Recommendations:

• Record object calibrations trials at a relatively low frame rate (20-30 Hz) to reduce file size and processing time.

• Object calibration trials only need to be a few seconds long.

Recording Object Calibrations:

1. Place your calibration object within the capture volume, at the desired position and orientation to define the global coordinate system.

2. Ensure the key points on your calibration object are visible in every camera view, and there are no obstructions.

3. Begin the recording.

4. After a few seconds, end the recording.

Shortcut: Ctrl + F4

Clear all loaded data and analysis results.

Shortcut: Ctrl + O

Shortcut: Ctrl + Shift + O

Open a previously saved workspace. To load the workspace, browse to and select the directory containing the workspace files (AVI, T3D, P3D).

Shortcut: Ctrl + S

Save the videos, camera calibration, and analysis results in the current workspace to a directory in a format that can be opened by Theia3D. Select the frames to include in the workspace and press the Save button. Browse to the desired save location and enter the name of the directory that will be created to store the workspace files. To overwrite the current workspace, use the dialog to select a .t3d file in the existing workspace folder. Note that this only overwites the results (T3D and P3D files) and leaves the videos untouched. In this case, all frames in the workspace must be saved. It is not possible to save to an existing workspace folder other than the currently open one.

‍

Save the current calibration. To save the calibration browse to the desired save location and enter the desired filename.

Save the scaled 3D model. To save the scaled model browse to the desired save location and enter the desired filename. Note that if multiple people are tracked a separate model will be saved for each of them. The files are automatically named as filename_personID.tmb, where personid is automatically assigned.

Shortcut: Ctrl + Shift + S, 1

For C3D files there is an option to Save multi-subject C3D. If this option is selected, two C3D files will be created - one with the unfiltered pose and one with the filtered pose. Both files will contain all tracked subjects. If the multi-subject option is not selected, separate files will be created for each tracked subject.

‍

‍

For JSON files the user can select which tracked individuals and which frames of the trial to include in the output file. The JSON output files are easily readable in a text editor or used as input for a custom analysis script.

‍

With the desired options selected, press the Save button, browse to the save location and enter the filename. The person ID and filtered tag are appended to the filename as needed.

Opens the pose data for all tracked people in Visual3D.

Shortcut: Ctrl + Shift + S, 2

Extrinsic chessboard calibration trials are used to determine the position and orientation of every camera in your camera system relative to your desired global coordinate system. As with lens calibration trials, users with wand calibration-capable systems (e.g. Qualisys Miqus, Vicon FLIR Blackfly S systems) are not required to complete chessboard calibration.

A minimum of one chessboard calibration trial should be collected every time you set up your camera system, but collecting multiple is recommended. A new calibration is required any time a camera is moved, so having multiple calibration trials collected throughout a long data collection session can help prevent data loss due to accidental or unnoticed camera movements.

Recommendations:

• Record chessboard calibration trials at a relatively low frame rate (20-30 Hz) to reduce file size and processing time.

• Keep the chessboard as flat as possible during the calibration trial.

• Focus on achieving groupings of 3+ cameras that can see the chessboard at all times.

• Align the chessboard with marks on the ground so you can confirm that the global coordinate system is positioned correctly when processing movement trials.

Avoid:

• Obstructions and occlusions; be aware of any objects within your capture volume which may partially or completely block the chessboard pattern from the view of the cameras. Even partial obstructions of the chessboard pattern may prevent it from being detected.

  • For treadmill capture volumes, be aware of treadmill safety rails, harnesses, etc.

  • For athletic environments, be aware of safety netting or equipment. Where possible, move or remove netting during calibration to ensure a direct view of the calibration chessboard.

• High frames or excessively long calibrations; chessboard calibration recordings produce the strongest result when many diverse perspectives of the chessboard are captured. Longer calibrations can provide more diverse data, but it is best to avoid high frame rates or excessively long calibrations to avoid long processing times.

Recording Chessboard Calibrations:

2. Choose or place visible marks on the ground which will be aligned with the global coordinate system, for confirmation when processing data.

3. Check that the chessboard is fully visible in at least 3 camera views when it is placed at the desired global coordinate system origin position.

4. Begin the recording.

5. Slowly ‘show’ the chessboard to groupings of 3 or more cameras while varying the position and orientation of the chessboard slightly. Ensure the chessboard is visible to cameras that overlap between groupings, ideally so that no fewer than 3 cameras can see the chessboard at all times. Focus on achieving groupings of 3+ cameras at all times.

6. When you are confident that all cameras have had sufficient views of the chessboard, slowly place the chessboard on the ground, aligned with your preselected marks.

7. Ensure you are not obstructing the view of the chessboard in any cameras during this localization phase.

8. End the recording.

1. 2D Views

The videos from each camera view are shown here and the 3D scene is projected onto each view. When all views cannot fit in the available area, arrows are shown and can be used to move through the available views. Double-click on a view to maximize it. When maximized, the mouse is used to zoom (scroll) and pan (right-click and drag) the view, and clicking on the arrows moves between views. Double-click the maximized view to restore it and show all views. Use Alt + right-click to inspect a view through a magnifying class - the level of magnification is controlled by the mouse wheel and reset with the ‘R’ key.

2. 3D View

The 3D scene is rendered here. The mouse is used to rotate (left-click and drag), pan (right-click and drag), zoom (scroll), and inspect (Alt + right-click) the scene. The 3D scene contains the global coordinate system, cameras, and skeletons of each identified and tracked person. The origin of the global coordinate system is a white sphere, the x-axis is a red arrow, the y-axis is a green arrow, and the z-axis is a blue arrow. The scene enviroment, ground, and lighting can be modified from the menu accessed using the button at the top left of the 3D view.

3. Quick Tools

The quick tools are shortcuts to the most common Theia3D commands. The entire processing pipeline can be executed using these shortcuts.

4. Menu

The menu is divided into eight sections.

5. Playback Controller

The video data can be played and paused using the play button on the left, or you can scroll through the trial using the timeline slider. The number to the right of the timeline indicates the current frame number, and the playback dropdown allows you to adjust the steps between frames that are displayed. When set to one, every frame is displayed, when set to two, every other frame is displayed, and so on.

‍6. Status Bar

Below the playback timeline controller is a status bar which indicates the current status and any ongoing processes.

The extrinsic camera calibration can be checked after the trial has been analyzed fully. For each view, the 2D representation of the skeleton in that view is compared to the 3D skeleton computed from all views. If the alignment can be improved in at least 30% of the frames by translating the entire skeleton within the 2D image for that view, this information is reported in the dialog. You can then choose to deactivate the cameras that are out of calibration and clear the entire analysis.

Graph Joint Angles

• Subject: The person to plot Euler angles for.

• Reference Segment & Segment: The Euler angles describe the orientation of the Segment relative to the Reference Segment, resolved in the Reference Segment coordinate system using a ZYX cardan sequence.

• Flip X/Y/Z: Flip the X, Y, or Z angle plots.

Show Batch Progress

View Command Results

The view command results window shows the commands that have completed by Theia3D while handling and processing data, including specific inputs and outputs in the dropdown menus when expanded. The text color indicates the step status, where green indicates success and yellow indicates a warning is present.

Show all will show hidden, non-critical commands in the list.

Clear Results will clear any commands currently in the Commands list.

Most of the Display options shown here can also be toggled ON and OFF using the Display toggle buttons in the sidebar.

Show/Hide 3D View

Toggle showing the 3D View in the application. If toggled ON but the 3D View is not visible, drag the 3D View open from the right border of the application window.

Person identification boxes can be toggled on and off using the sidebar button.

Toggle the boxes around all people found in each 2D view. The color of the box around each identified person is unique to that person and the same in all views. The boxes around people who have not been identified are grey. Person identifications are also printed in the upper left corner of the boxes.

3D segments can be toggled on and off using the sidebar button.

Toggle the 3D segments of each identified person in the 2D views and in the 3D view. The color of the segments is unique to the identifed person and matches the color of the 2D boxes corresponding to that person.

Skeletons can be toggled on and off using the sidebar button.

Toggle the skeletons of each identified person in the 2D views and in the 3D view. The color of each skeleton is unique to the identified person and matches the color of the 2D boxes corresponding to that person.

Local segment coordinate systems can be toggled on and off using the sidebar button.

Toggle the local coordinate systems of segments and cameras in the 2D views and in the 3D view. Note that the local coordinate systems of the segments are only shown if the segments or skeleton are visible. The origin of each coordinate system is a white sphere, the x-axis is a red arrow, the y-axis is a green arrow, and the z-axis is a blue arrow.

Cancel the currently running analysis. This option is available (and replaces all other options) only when analysis is running.

Detect people in each of the loaded videos. When complete, boxes are drawn around the detected people in the video overlays if Show Boxes is selected in the Display menu. Video data and calibration must be loaded before running Run 2D.

Identify people across views (i.e., determine “who is who” in each of the views). Identification requires the person to be clearly visible in at least three views simultaneously. Each identified person is assigned a unique color that is applied to their boxes, segments, and skeleton. Run 2D must be completed before running Track People.

Runs: Track People and Solve Skeleton. Run 2D must be completed.

Shortcut: Ctrl + F

Runs: Run 2D, Track People, and Solve Skeleton. Video data and calibration must be loaded.

Synchronization of the video images can be checked after the trial has been analyzed fully. For each view, the 2D representation of the skeleton in that view is compared to the 3D skeleton computed from all views. If the alignment for a given view can be improved by shifting its video sequence forward or backward in time, this frame offset is reported in the dialog. You can then choose to correct the synchronization by applying these offsets to the out-of-sync views and clearing the entire analysis.

The Adjust Calibration tool can be used to modify the position and orientation of the global coordinate system (GCS) after completing a chessboard or object calibration, or after loading an existing calibration file. The GCS projection on each 2D camera view and within the 3D View is updated live as the Position and Angle sliders are used to modify these parameters of the GCS localization.

This tool can be used to reposition the GCS to a more desirable position or orientation if the desired origin frame could not be used from the chessboard calibration trial, for any other methodological reason that requires the GCS in a specific position.

Position (X, Y, Z)

X, Y, Z sliders can be used to change the position of the GCS relative to its original position, measured in mm.

Angle (X, Y, Z)

X, Y, Z sliders can be used to change the orientation of the GCS relative to its original orientation, about the respective axis of the GCS, in degrees.

Reset

Reset the position and angle sliders to 0.0.

Apply

Apply the selected position and angle adjustments to the current camera system calibration.

Apply and Save

Apply the selected position and angle adjustments to the current camera system calibration, and save the result as a new calibration .txt file.

Load Videos

Frame Grab Step

The step between frames searched for chessboards. Smaller values take longer to process, but may provide slightly improved calibration results compared to larger values.

Camera Type

Detects and indicates the camera manufacturer based on the videos loaded. This allows the automatic application of default Sony RX0 II camera intrinsic parameters when the calibration videos were recorded using this camera system (see below).

Use Default RX0 II Intrinsics

Automatically selected after loading videos recorded using a Sony RX0 II camera system. This option should be selected to utilize the built-in intrinsic parameters for the Sony RX0 II cameras. Note: These parameters are only valid for Sony RX0 II cameras with standard lenses. For any other cameras or lenses these default parameters are invalid.

‍Load Custom Intrinsics

Loads intrinsic lens parameters for the camera views loaded.

Origin Frame

Frame to use to set the global coordinate system of the capture volume. If the chessboard is not detected in at least three views in this frame, the closest frame in which it can be detected will be used.

Min 2 Cams for Origin Triangulation

Uses two (instead of the default of three) cameras to locate the origin frame. Use this option if the board is difficult to see in the triangulation frame from three cameras.

Normal Axis

Set the coordinate system axis that is defined by the normal axis of the chessboard.

Long Axis

Set the coordinate system axis that is defined by the long axis of the chessboard.

Use Custom Chessboard

Use a custom size chessboard for the calibration. Only select this option if not using the chessboard provided by Theia Markerless.

• Square Size: The width and height of each square on the chessboard. Measured in mm.

• Number of Squares High: The number of inner squares in the vertical direction of the chessboard.

• Number of Squares Wide: The number of inner squares in the horizontal direction of the chessboard.

Calibrate Cameras

Perform the extrinsic calibration for all of the cameras.

Chessboard Calibration Metrics

After the camera calibration is complete, a dialog will appear with the results of the chessboard calibration trial. The result metrics can be interpreted as follows:

Frames

The number of video frames used to calibrate each camera.

RMSE Reprojection

RMSE error of the reprojected 3D chessboard points relative to the detected chessboard points in 2D, for each camera view. Measured in mm. Since this is an error relative to 0 mm for all chessboards regardless of size, a 1 mm RMSE reprojection has the same meaning for all sizes of chessboard.

RMSE Diagonal

RMSE error of the length of the diagonals of the 4 outer corners of the chessboard relative to their known lengths, for all triangulated chessboard detections. Measured in mm. Provides an absolute error measure, not normalized to the size of the chessboard. Therefore, a 1 mm RMSE diagonal for a large board is a smaller percentage error than a 1 mm RMSE diagonal for a small board. Not affected by the number or size of the chessboard squares. Recommend using <1 mm as ‘Excellent’ and <2 mm as ‘Acceptable’; calibrations with RMSE Diagonal above 2 mm are not recommended for use.

RMSE Angle

RMSE error of the angle of the 4 outer corners of the chessboard relative to their known angle of 90 degrees, for all triangulated chessboard detections. Measured in degrees. Since this is an error relative to 90 degrees for all chessboards regardless of size, a 1 degree RMSE angle has the same meaning for all sizes of chessboard. Not affected by the number or size of the chessboard squares.

RMSE Flat

RMSE error of the normal distance from each detected chessboard point in 3D to the flat plane formed by the outer four 3D chessboard points. Measured in mm. Provides an absolute error measure, not normalized to the size of the chessboard. Therefore, a 1 mm RMSE flat for a large board indicates a lesser degree of bend in the chessboard than a 1 mm RMSE flat for a small board.

Origin Triangulation Frame

The video frame used to initialize the position and orientation of the global coordinate system.

After reviewing the calibrations results, the folling options are available:

Save allows the calibration to be saved.

Save & Assign initiates two steps:

1. Opens the Save Calibration window, allowing the calibration to be saved as a .txt file.

2. Opens the Assign Calibration tool, allowing the previously saved calibration .txt file to be immediately assigned to movement trials.

Ok acknowledges and closes the results dialog window.

Chessboard Calibration Review

After the camera calibration is complete, the calibration trial videos can be reviewed for feedback. The following visual cues projected onto the camera views can be useful in determining which portions of the trial contributed to the calibration, and those that did not. This can be useful for optimizing your calibration trial technique.

Load Videos

Camera Type

Detects and indicates the camera manufacturer based on the videos loaded. This allows the automatic application of default Sony RX0 II camera intrinsic parameters when the calibration videos were recorded using this camera system (see below).

Load Default RX0 II Intrinsics

Load default lens calibration parameters for Sony RX0 II cameras and apply them to all cameras. This can be used to avoid performing a lens calibration when using the RX0 II cameras. Loading default intrinsics overwrites any existing lens calibration parameters. Note: These parameters are only valid for Sony RX0 II cameras with standard lenses. For any other cameras or lenses these default parameters are invalid.

Load Object

Load a TXT file defining the calibration object. Each line of the file contains comma-separated x-y-z coordinates of one of the key points on the object in mm. To load the object, browse to and select the object file. Once loaded, the object points will be displayed in the table.

Save Object

Save the current object definition. To save the object, browse to the desired save location and enter the desired filename.

Add

Add a calibration object point definition (x, y, z coordinate values).

Remove

Add a calibration object point definition (x, y, z coordinate values).

Auto-Step

If checked, the selected object point will advance each time a point is identified in one of the 2D views.

Cam ID

The ID of the camera corresponding to the current 2D view. Use the drop-down to select a different 2D view. Maximizing a view will automatically select it.

Reset Point

Remove the currently selected point from the current view.

Reset Camera

Remove all points from the current view.

Reset All Cameras

Remove all points from all views.

Calibrate Cameras

Perform the extrinsic calibration for all of the cameras. After performing the calibration, the position of any camera points that were not identified before the calibration are calculated and drawn in the 2D views as an ‘*’.

Adjust Origin

When using the checkerboard extrinsic calibration, sometimes the origin may appear at an undefined location because the board was not recognized in the origin frame. This feature allows you to adjust the reference frame by clicking on an object (typically four squares on the checkerboard). Since the frame is just being adjusted (not computed), a calibration needs to be loaded to use this feature, and the relative pose of the cameras will be unchanged.

The Organize Videos tool converts a folder of videos into a structure that can be used by Theia3D. This functionality is useful when collecting large data sets.

Basic

The Basic tab is the best option when organizing videos that have user-defined file names, such as those collected using OptiTrack, Qualisys, or Vicon camera systems. To use this tool, it is recommended to include relevant details about the trial as part of the file name, consistently separated by one delimiter from the following options:

• Space (' ')

• Hyphen ('-')

• Underscore ('_')

• Period ('.')

• Space (' ') and Underscore ('_')

With video files named in this way, the dropdown options below can be assigned to each delimited portion of the filename, resulting in an intuitive file structure that maintains this information for each trial:

• Cam ID: (Required) The ID of the camera corresponding to the video file. This must be the same ID as in the calibration file.

• Subject: (Recommended) Unique subject identifier (e.g. sub001).

• Action: (Recommended) The action the subject is performing (e.g. walk).

• Trial: (Recommended) The trial number (e.g. 001 or walk01).

• N/A: (Optional) Additional terms in the filename not used by Theia3D.

To format video data using this tool, follow the steps below:

1. Select the delimiter used in the file name.

2. Use the Browse button to select the folder containing the videos. Once the directory has been selected, the dialog will parse the name of a sample file. Note that the delimiter can be changed after the folder is selected and the sample file name will be re-parsed.

3. Use the drop down menus to select the reserved term that each part of the filename corresponds to.

4. Click Execute to organize the data into the required structure. The hierarchy of the created directory is Subject > Action > Trial > Cam ID. Cam ID is the only reserved term that must be included in the filename; however, it is recommended to include any other relevant details about the trial as part of the filename.

For example, consider a folder with the following video names:

• S0001_Walk_001_8_21375.avi

• S0001_Walk_001_8_21376.avi

• S0002_Walk_001_8_21375.avi

To format these videos, the reserved terms selected are: Subject, Action, Trial, N/A, and Cam ID. The data structure after formatting is:

Top Level Only

When enabled, this option sorts only the video files located directly within the current folder. Any subfolders and the files they contain are ignored during the sorting process. This ensures that only the top-level contents are affected, and can be useful when adding new, unorganized data into a folder with existing, organized data.

Advanced

The Advanced tab is the best option when organizing videos that do not have user-defined file names, such as those collected using Sony RX0 II camera systems. To use this feature, the names of the video files must end with camid_trialid, where camid is the ID of the camera that recorded the video and trialid is a unique identifier for the trial.

‍When exporting videos using the Sony RX0 II camera system browser interface, make sure that the Name of Imported File options are selected as follows:

• Add the Camera Label (prefix)

• Do not add the Shooting Date/Time

To format video data using this tool, follow the steps below:

1. Use the Browse button to select the folder containing the videos. The trials table is then created with one row for each trial. The File column of each rorow is filled with the trialid for that trial.

2. Use the Load Subject and Action Assignments button to load the subject and action IDs for each of the trials from a csv file. Each row of the csv file must contain the trialid, subjectid, and actionid of a single trial, in that order, separated by commas. For example, the csv file used to assign the subject and action IDs in the above figure was:

C0001,Calibration,Chessboard
C0002,S0001,Walk
C0003,S0002,Walk

1. Press the Execute button to format the videos in the required folder structure.

Alternatively, Subject and Action details can be manually assigned to the trials table without using a CSV file:

1. Double click any cell to type in the Subject or Action ID.

2. Click Save Subject and Action Assignments to save the filled table as a CSV.

3. Press Execute to format the videos in the required folder structure.

To use this tool, the raw video data must be recorded using the Sony RX0 II camera system, and should be downloaded using the options: Add the Camera Label (prefix) and Do not add the Shooting Date/Time. The raw video files must be accompanied by a data specficic CSV file, which can have any name, however it must be the only .csv file within the directory alongside the video files. The data spec .csv file must have a header row of: Trial,SubjectId,Type,Action,CalibrationId. Each subsequent row should contain the relevant information under each header column that corresponds to the trial.

For example, a data collection consisting of the following recordings in this order:

• subject 1, walk

• subject 1, run

• subject 2, walk

• subject 2, run

• calibration

• subject 3, walk

• subject 3, run

would require a data specific CSV as follows:

After successful formatting of the data, an additional file changelog.csv is created which documents the mapping of raw video files to the organized file structure.

The assign calibration file tool is used to copy and nest an extrinsic calibration file within individual trial folders, one level above the video files. This is best practice for storing calibration files with their associated movement trials, and is a requirement for running batch analyses.

Path to Calibration File

Use the Browse button to navigate to and select your extrinsic calibration .txt file.

Plus‍

Use the plus button to add trials to the Trials list by selecting a folder containing movement trials. All trials within the selected folder will be added to the Trials list, regardless of their depth within the selected folder.

‍

Minus‍

Use the minus button to remove currently selected trials from the Trials list. Use shift+click or ctrl+click to select groups or individual trials to remove from the Trials list.

Just Add to Selection

Assign the calibration file to only the currently selected files within the Trials list.

Cancel

Close the Assign Calibration Files tool without assigning any calibration files.

Assign

Assign the calibration file to the trials in the Trials list or only the currently selected trials if Just Add to Selection is selected.

The Display Video Metadata tool can be used to easily review relevant metadata for the videos currently loaded, including:

Camera Id

Camera ID dropdown, allowing separate cameras to be selected and reviewed.

Video Path

Path to the video file for the currently selected camera.

Trial

Name of the lowest folder level.

Size

Video image resolution, in pixels.

Number of Frames

Video file length, in frames.

Frame Rate

Video file frame rate, in frames per second.

Frame Offset

Number of frames offset between the current video file and other synchronized video files from the current recording. Generally a value of 0 is expected for most camera systems that record frame synchronized videos using on a start/stop trigger signal. Sony RX0 II camera systems may have non-zero values here, as these systems use timecode video synchronization for post-hoc alignment of videos between cameras.

Load Videos

Frame Grab Step

The step between frames searched for chessboards. Smaller values require longer to process, but may provide improved calibration results compared to larger values.

Use Custom Chessboard

Use a custom size chessboard for the calibration. Only select this option if not using the chessboard provided by Theia Markerless.

• Square Size: The width and height of each square on the chessboard. Measured in mm.

• Number of Squares High: The number of inner squares in the vertical direction of the chessboard.

• Number of Squares Wide: The number of inner squares in the horizontal direction of the chessboard.

Calibrate Lenses

Perform the intrinsic calibration for all of the cameras.

Save Intrinsics

Save the intrinsic calibration parameters in a format that can be loaded during future calibrations. This is useful when the intrinsic parameters do not change between calibrations. To save the intrinsic calibration, browse to the desired save location and enter the desired filename.

Load Intrinsics

Load the intrinsic calibration parameters from a previous calibration. This is useful when the intrinsic parameters do not change between calibrations. To load the intrinsic calibration, browse to and select the previously saved intrinsic lens calibration file.

Merge Intrinsics

Merge the intrinsic calibration parameters from the current calibration trial with those from a previous calibration. This is useful when new cameras have been acquired that need to be added to an existing intrinsic calibration file, or if a new intrinsic calibration trial has been recorded for a subset of cameras that are already in an existing intrinsic calibration file and the old camera parameters should be replaced. To add or replace intrinsic parameters for a set of cameras within an existing intrinsic calibration file, load the intrinsic calibration trial for those cameras, select the Frame Grab Step, and click Calibrate Lenses.

Lens Calibration Metrics

After the lenses have been calibrated, click Merge Intrinsics, then navigate to and open the existing intrinsic calibration file to which the new parameters should be merged. A new, merged intrinsic calibration file will be saved next to the existing file.

After the lens calibration is complete, a dialog will appear with the results of the intrinsic lens calibration trial.

‍The result metrics can be interpreted as follows:

Chessboards

The number of video frames used to calibrate each camera.

Coverage

The proportion (maximum value 1) of the camera view covered during the intrinsic lens calibration trial. Recommend >0.9 as a quality threshold.

Angle

The maximum angle of the chessboard relative to the camera image plane during the intrinsic lens calibration trial. Recommend 30-60 degress as a quality threshold.

Lens Calibration Review

After acknowledging the lens calibration results dialog window, the camera views will be updated to visualize the results of the lens calibration process. The green shading is a heatmap of the detected chessboard across the camera view, where the green area indicates the portion of the view that was covered during the calibration and the intensity of the green area indicates the number of frames in which the chessboard was detected while covering that part of the view.

The Video Enhancement tool allows you to improve video quality before processing data, which can be especially useful for calibration under challenging lighting conditions. Changes made using this tool are previewed for the loaded 2D views, and any enhancements are applied before running the 2D inference, lens calibration, or chessboard calibration processes. Therefore, what you see in the 2D views is how the video data will be processed.

All Views

Apply the selected enhancements to all camera views. Alternatively, select an individual camera from the dropdown menu to apply individual camera enhancements.

Brightness

Increase or decrease the brightness of the videos. Default (no enhancement) is 0.

Contrast

Increase the contrast of the videos. Default (no enhancement) is 1.00.

White Balance

Adjust the white balance of the videos by performing a color correction using a baseline white color. Enable White Balance by selecting the square, and click on the white rectangle to open the color picker. You can select a specific color, input color values, or use the Pick Screen Color option to select a pixel from within your loaded videos of a white surface such as the chessboard.

Blue Mask

Shows which pixels are detected as blue by the chessboard calibration algorithm. This is useful for adjusting your videos before processing a chessboard calibration to ensure the blue squares are detected. The global coordinate system origin is defined at the intersection of the blue squares, which must be detected in order to define the origin.

Reset

Reset all changes to default values, removing any video enhancements.

Shortcut: Ctrl + ,

‍

The settings available at the top of the preferences pane enable the user to load, save, and save default preferences.

Loaded Preferences File

Displays the path to the preferences file from which the current preferences were loaded.

• Load: Allows the selection of a new preferences .pxt file.

• Save As: Allows the current preferences to be saved as a new .pxt file.

Default Preferences

• Load: Allows previously specified default preferences to be selected and loaded.

• Save As Default: Allows the current preferences to be saved as a new set of default preferences.

The Analysis preferences pane contains advanced options and parameters for adjusting how the movement is analyzed, including subject identification and person tracking, model parameters, and 3D reconstruction parameters.

Analysis Frame Range

First and last frame to analyze (inclusive). Modified frame range values are included when preferences are saved. (Default: full trial length)

Image Matte Percentage

Image exclusion border size. Image matte area is displayed as a grey border around the camera views when active. The greyed-out portions of the videos will not be used when running the analysis, allowing poorly calibrated image borders to be ignored. Image matte area is not applied during lens calibration.

Max People

Maximum number of people to track. (Default: No max)

Person Tracking Mode

Select the method for determining person tracking priority. Run Analysis (without 2D) must be performed in order to update person tracking.

• Most Visible: Person identification and tracking performed by prioritizing people who are visible in >75% of the total frames from all cameras throughout the entire trial, and ordering them by their distance to the global coordinate system origin.

• Closest to Origin: Person identification and tracking performed by ordering all people by their minimum distance to the global coordinate system origin throughout the entire trial. For example, if a person walks directly over the origin, they will likely be identified as person 0.

• Order of Appearance: Person identification and tracking performed by numbering people in order of appearance within the capture volume.

Remove stale IDs

Remove any person IDs that are missing for 100 consecutive frames or more. This allows a person who is tracked repeatedly within a single trial (e.g. during several passes into and out of the volume) to be tracked and therefore exported separately, for instance as multiple separate c3d files.

Track Rotating People (beta)

Improve tracking of rotating and non-upright people at the expense of speed. (Default: Off)

If not selected, the generic model will be scaled during Solve Skeleton. If selected, a previously saved model can be used by the inverse kinematics algorithm. Use the Use Saved Model button on the next line to select the model to be used. Note that this option is only available when Max People is 1. (Default: Off)

Large Lens Distortion

Uses a larger grid for lens distortion correction. Only use if there is a lot of distortion and loading is slow. (Default: Off)

Default Model (neither Full Body nor Separate Arm and Head model selected)

‍Use Full Body Model

‍Use Separate Arm and Head Models

Enable 3 DOF Knee

If selected, the model used by the inverse kinematics algorithm will have three degrees of freedom at the knee (flexion/extension, ab/adduction, and internal/external rotation). If not selected, the model will have two degrees of freedom at the knee (flexion/extension and ab/adduction). (Default: Off)

Smoothing Freq. (Hz)

The cutoff frequency of the lowpass filter used to smooth the pose from the inverse kinematics. The default GCVSPL cutoff frequency is 20 Hz, cutoff frequency provides some smoothing and noise reduction while generally being sufficiently high to avoid ‘over-smoothing’ most movements.

General Recommendations:

• For slow movements (e.g. sway, slow walking), a lower cutoff frequency around 6–12 Hz can be used.

• For faster movements (e.g. running or jumping), a higher cutoff frequency around 10–20 Hz may be required.

• Movements with very high segment velocities or accelerations such as baseball pitching or very dynamic changes of direction, cutoff frequencies above 20 Hz may be required.

Show Warning

When applying the filter using the selected smoothing frequency, Theia3D checks whether the applied smoothing frequency is appropriate for the movement and warns the user if they should consider adjusting the frequency. This test can occasionally produce false positive warnings, so these warnings can be disabled if desired.

Display/Use

Enables the 3D Analysis Bounding Box for use, which restricts person tracking to within a specified 3D volume. This allows people who are visible but not of interested to be ignored, such as the experimenter or other observers. Run Analysis (without 2D) must be performed in order to update person tracking and skeleton visibility.

Use Camera Locations

The 3D camera locations will be used to establish the edges of the analysis bounding box, rather than the x, y, and z origin positions and length, width, and height dimension definitions.

• Origin X: Position of the center of the bounding box along the global coordinate system X-axis.

• Origin Y: Position of the center of the bounding box along the global coordinate system Y-axis.

• Origin Z: Position of the center of the bounding box along the global coordinate system Z-axis. Minimum value is half the bounding box height.

• Length: Bounding box dimension along the global coordinate system X-axis.

• Width: Bounding box dimension along the global coordinate system Y-axis.

• Height: Bounding box dimension along the global coordinate system Z-axis.

‍The Toggle Views tool can be used to turn on and off individual camera views for the currently loaded trial. When turned off, camera views are greyed out in the 2D viewer area, and are not used when analyzing the trial.

Toggle Views adjustments are included when saving preferences files.

Opens the online documentation for the software.

Submit Support Ticket

The About dialog shows the following information:

• Software version number

• Release date

• License key

• License counts

• Support expiry

• GPU utilization information

‍

‍

The Setup preferences pane contains options and parameters for the software setup and startup.

Path to Visual3D.exe location. Must be set in order to load data in Visual3D from Theia3D. Use the Visual3D Path button to browse to and select your Visual3D.exe file.

Select GPUs on launch.

Show or hide the startup/GPU selection dialog when Theia3D launches.

The Modify People IDs tool allows you to swap the unique person ID numbers assigned to each person when multiple people are tracked in a trial. Click and drag an ID and drop it onto another ID to swap those two ID numbers. This change to the IDs is immediately visible in the 2D and 3D views. You can also right-click on an ID to display a context menu with options for removing that tracked person or making it person 0.

The rendering preferences pane contains advanced options and parameters for how the data should be played back and rendered.

Playback

Playback rate during continuous playback. A value of 1 will show every frame, 2 will show every other frame, etc. (Default: 1)

Skeleton Alpha

Set the opacity of the skeletons, 3D segments, and local coordinate systems rendered in the 2D views. The slider ranges from 0 (transparent) to 100 (opaque).

Render Smooth IK

If selected, the filtered pose of the skeletons will be rendered. If not selected, the unfiltered pose will be rendered. (Default: Off)

The full body kinematic model consists of one kinematic chain comprised of the lower body, upper body, and head, with the pelvis as the root segment. Abdomen and neck segments are included. The Full Body Model must be used in order to save skeleton poses as FBX format.

Full Body Kinematic Chain

There are currently three available models in Theia3D:

1. • Consists of two kinematic chains: lower body (pelvis and legs) and upper body (torso, arms, and head).

   • No abdomen or neck segments.

   • Shoulder joints and head segment are allowed 6 degrees of freedom.

2. 1. Consists of one, whole-body kinematic chain.

   2. Abdomen and neck segments included.

   3. Shoulder joints are allowed 6 degrees of freedom.

   4. Head is allowed 3 degrees of freedom.

3. • Consists of five kinematic chains: lower body (pelvis and legs), torso, left arm, right arm, and head.

   • No abdomen or neck segments.

   • Shoulder joints and head segment are allowed 6 degrees of freedom.

Pose

Model pose can be exported to .c3d, .fbx, and .json files.

The .c3d files contain the 4x4 pose matrices for each model segment and the local coordinates of the anatomical landmarks of the distal segments of the model (feet, hands, head). These files can be processed using Visual3D.

The .fbx files contain the hierarchical skeleton, pose, and bone meshes of the animation model. The first frame of the file contains the model in a “T-Pose”. The skeleton must be solved using the Full Body Model in order to save pose files in FBX format.

The .json files contain information about how the trial was processed (Theia3D version, model, preferences, etc.) and the 4x4 pose matrices for each body segment for every frame of the trial.

Inertial Properties

With each model, inertial properties are estimated using established regression equations similar to Dempster’s, which provide values for segment masses, centers of mass, and moments of inertia based on subject-specific anthropometric estimates and validated population data. Theia3D combines deep learning with inverse kinematics to estimate 3D body movement, using calculated segment positions, orientations to support further biomechanical analysis.

Coordinate System

The 4x4 pose matrices for each body segment convey the position and orientation of the segment’s local coordinate system, which are visualized with the x-axis as red, y-axis as green, and z-axis as blue. Local coordinate system origins are visualized as white spheres

The Default kinematic model consists of two kinematic chains for the lower body (pelvis and legs) and upper body (torso, arms, and head).

Upper Body Kinematic Chain

Lower Body Kinematic Chain

Kinematic Model Tables

Torso Kinematic Chain

Head Kinematic Chain

Right Arm Kinematic Chain

Left Arm Kinematic Chain

Lower Body Kinematic Chain

A workspace folder contains the video and data files of a saved workspace. The video files are named according to their unique camera ID. The .t3d and .p3d files contain the analysis results. It is important that the contents of the workspace folder are not modified, including moving files in or out of the folder.

Theia3D workspaces can be used to save data at various stages of analysis, but are most useful for saving analyzed data that will be reviewed later. When a fully analyzed movement trial is saved as a Theia3D workspace, the saved data includes the 2D videos, calibration file, and all data associated with the analysis. Therefore, when a saved workspace is loaded, you can immediately review the analyzed data in the 2D and 3D viewers.

Top open a Theia3D workspace while Theia3D is closed, you can double-click on the results P3D file.

Video Data

‍

The video data for a single trial must be contained in its own folder, and each video file must be in its own subfolder. The name of each subfolder must be the ID of the corresponding camera. There are no requirements on the structure of the names of the video files, but the names must be unique and videos must be .avi or .mp4 format. For example, the figure above shows video data for a walking trial collected using four cameras with IDs 21375, 21379, 21380, and 21381.

FBX file outputs can be saved with a variety of different segment coordinate system conventions for your convenience.

Output JSON files contain:

• Theia3D version number, engines, and kinematic model information

• Trial frame rate

• Preferences used during process, such as start/end frame, max people, smoothing frequency, etc.

• Tracking data for each tracked individual, including ID number, segment names and 4x4 pose matrices, and parameters such as segment length.

To save video overlays from Theia3D, use the Save Video Overlay button in the sidebar or under the File dropdown menu.

To save video overlays from Theia3D Batch, select Save Videos and any desired display options and camera views to be saved, prior to running the analysis.

One of the benefits offered by Theia3D markerless motion capture is automated tracking, which allows it to analyze large datasets without human intervention or supervision. This is achieved using the Theia3D Batch companion application to Theia3D, which allows a list of trials to be curated and batch analyzed sequentially. While batch processing is efficient and does not require supervision, we always recommend that you manually examine and check the quality of your markerless data and calibrations using Theia3D, before setting up a batch analysis. This can prevent poor calibrations or other issues with the data from going unnoticed until after the batch analysis has been completed.

Camera Calibration File

‍

The camera calibration file contains the calibrations for all of the cameras using a structure similar to XML. An example calibration file is provided with the sample data. Key elements and attributes of the calibration file are outlined here. Note that Qualisys calibrations exported from QTM as TXT files and Vicon calibrations (XCP) can also be used.

‍<calibration>

Top level element with no attributes.

Grouping element that holds <camera> child elements. It has no attributes.

<results>

<camera>

Element that holds the calibration information for a single camera. Required attributes are: active, serial, and viewrotation. Required child elements are: <transform> and <intrinsic>.

• active: 1 if the camera is used, 0 if unused.

• serial: The camera ID. Must be unique to the camera and match the camera ID used to name the video subfolders.

• viewrotation: Rotation of the camera. 0 if upright, 180 if upside down, 90 or 270 if sideways. This is calculated from the rotation matrix of the calibration.

<transform>

Translation and rotation components of the transformation from global to camera coordinates (p→cam=Rp→global+t→). Required attributes are x, y, z and r11 through r33.

• x, y, z: Elements of the translation component (t→) of the transformation. Expressed in mm.

• rij: Elements of row i and column j of the rotation component (R) of the transformation.

<intrinsic>

Intrinsic camera parameters. Required attributes are: focallength, sensorMinU, sensorMaxU, sensorMinV, sensorMaxV, focalLengthU, focalLengthV, centerPointU, centerPointV, skew, radialDistortion1-3, and tengentialDistortion1-2.

• focallength: Focal length of the lens in mm.

• sensorMinU: Minimum u coordinate of the sensor in pixels.

• sensorMaxU: Maximum u coordinate of the sensor in pixels.

• sensorMinV: Minimum v coordinate of the sensor in pixels.

• sensorMaxV: Maximum v coordinate of the sensor in pixels.

• focalLengthU: Focal length along the u axis in pixels.

• focalLengthV: Focal length along the v axis in pixels.

• centerPointU: Principal point u coordinate in pixels.

• centerPointV: Principal point v coordinate in pixels.

• skew: Skew coefficient. Non-zero if the image axes are not perpendicular. Note that non-zero skew is currently unsupported.

• radialDistortion1-3: Radial distortion coefficients.

• tangentialDistortion1-2: Tangential distortion coefficients.

Intrinsic vs. Extrinsic Calibration

Accurate 3D reconstruction in markerless motion capture requires both intrinsic and extrinsic calibration of all cameras in the system. These two types of calibration serve different but complementary purposes. The same calibration chessboard is used for both the Intrinsic and Extrinsic calibration.

Intrinsic Calibration (Lens Calibration)

Intrinsic calibration characterizes the internal properties of each camera, including:

• Focal length

• Sensor center point

• Skew

• Lens distortion

Extrinsic Calibration (System Calibration)

Extrinsic calibration defines the position and orientation of each camera within the global 3D coordinate system. This is essential for aligning multiple cameras into a unified space so they can work together to triangulate 3D points.

Extrinsic calibration requires:

• Chessboard

Single-camera views are insufficient to determine the board’s 3D position and orientation, therefore multi-camera visibility is critical and the best results are obtained when three or more cameras can see the chessboard pattern throughout the calibration trial.

Why Both Are Required

• Intrinsic calibration ensures each camera produces geometrically correct images by accounting for internal optical characteristics and enables projection from 2D pixels coordinates to 3D rays.

• Extrinsic calibration aligns the position and orientation of each camera relative to a global coordinate system. It requires multiple cameras to view the chessboard simultaneously to resolve the calibration board's 3D pose.

‍

The Open into options allow you to easily open the currently selected batch processing directory, in either Windows Explorer or the command line.

Convert Workspaces

When using this tool, two file selection dialog windows will appear, in the following order:

1. Workspace Directory: Use this dialog window to select a folder containing one or more Theia workspaces to be converted to raw data.

2. Output Raw Directory: Use this dialog window to select an empty folder where the extracted raw data should be output.

Once the above two dialog windows have been used, Theia3D Batch will automatically convert each workspace and output the raw data in the selected folder.

Move C3D Files

The Move C3D Files tool can be used after completing a batch analysis to copy and move output C3D files from within the lowest-level trial folders in the '_c3d' output folder to directly within the top-level folder of the '_c3d' output folder. The original C3D output files located in the lowest-level trial folders will not be deleted.

Settings

Data Path

Currently selected batch analysis root folder.

Browse‍

Select the batch analysis root folder.

Save Workspace

Save JSON

Save C3D

Save FBX

Save Video