This guide is provided as reference and educational material explaining how to use X Factory, and is made available for anyone to view. All information and content in this document are the property of Vueron Technology and may not be reproduced or modified without the written permission of Vueron Technology. Excerpts of this content may not be used without authorization. Vueron Technology makes no express or implied warranty regarding the content of this document, and the information provided is subject to change without notice. Revised editions may be issued when changes occur. The software described in this guide is provided under a license agreement or a non-disclosure agreement, and any action such as reverse engineering is prohibited except where expressly permitted by that agreement.

Getting started with X Factory

There are two ways to start using X Factory. You can try it first with a Free Trial, or purchase it and start right away.

Start with the Free Trial

You can start the Free Trial by signing up with your company email.

On the login screen, click Sign up.

Enter your company email address and click Confirm to check whether the email is available.

Enter your profile (name, company, job title) and workspace information.

Agree to the required terms and click Create account.

Click the verification link sent to your email to submit your sign-up request. The verification link expires after 10 minutes, and you can resend it from the screen.

Once your sign-up request is approved, a temporary password is sent to your registered email. Log in with this password to start using X Factory.

For the scope and duration of the Free Trial, see Understanding the Free Trial.

Start by purchasing

To start by purchasing, you need an AWS account. On the AWS page, log in to your AWS account, search for VueX (X Factory) on AWS Marketplace, and click the Subscription button to start using it.

Once your sign-up request is complete, your account manager will provide information such as a product introduction, contract guidance, and how to use X Factory through your registered email.

Understanding the Free Trial

This section explains the scope and duration of the X Factory Free Trial.

Usage limits

Item	Amount
AI annotation	2,000 runs total
Data capacity	50 GB

AI annotation is deducted by the number of runs. You can run AI annotation multiple times on a single frame, and one run is deducted each time. For example, if you run it ten times on one frame, you have used 10 of the 2,000 runs. In the extreme case, you could use all 2,000 runs on a single frame.

In other words, whether you run it once each on 2,000 different frames or several times on a few frames, up to 2,000 AI annotation runs in total are provided free of charge.

You can check your remaining runs when you run AI annotation, or on the Admin Setting page.

When you have used up your Free Trial limits (2,000 AI annotation runs or 50 GB of data capacity), you can upgrade to a paid plan to continue.

Manually editing labels created by AI does not consume runs. Runs are used only when you start a new AI annotation.

Trial duration

The Free Trial period starts from your approval date.

Full-use period (3 months): For 3 months from the approval date, you can use all features such as upload, AI annotation, and editing. However, the AI Modeling and Deployment features are not available; they can be used through a separate inquiry.

View-only period (3 months): After the full-use period ends, your data is kept in view-only (Read Only) mode for the next 3 months. During this period you cannot upload or edit, but you can view and download your existing data, or convert that workspace's data to a paid plan as-is.

Deletion: After the view-only period ends, the workspace data is deleted. Please download any data you need before the view-only period ends.

Collection (Data upload)

To get started with X Factory, you need to upload LiDAR data. X Factory supports the PCD format (*.pcd). Click New folder to create a folder to upload data into, and inside that folder click Upload or drag and drop to upload your data.

File upload rules

Each uploaded file must be 50 MB or less. (The upload file size can be customized; if you need a larger size, please contact us.)

You can upload folders, .pcd, .json, and image files (.png, .jpg, .webp).

.zip files are not supported. Please unzip before uploading. (A compressed file can be uploaded, but its contents cannot be read.)

When uploading a folder, the folder must contain a .pcd file.

Files that have already been uploaded will be overwritten.

Formats that are not LiDAR point clouds (PCD), such as .bin, are not supported for direct upload. If you would like to start with such data, please contact xfactory@vueron.ai.

Annotation

To train an AI model, the uploaded data needs to be processed. X Factory uses AI to automatically and quickly label the uploaded data.

(1) Create a project

Click Create project to create a project. Project Name and Member preset are required, and you can select the Object classes you want to annotate.

(2) Create a task

Select the project you created and click Create Task to select the data uploaded in Collection and assign it to the task.

Task Folder loads the folder specified in Collection. If you need to change it, click Select Folder.

Task Name defaults to the folder name and can be changed.

Start/Due Date can be set according to the task schedule.

Frame interval refers to the interval of frames you want to use.

Target Object defaults to the Object class selected when creating the project and can be changed.

Folder selection rule when creating a task

When creating a task, you must select a folder that contains .pcd files. Only the .pcd files directly inside the selected folder are loaded; files inside sub-folders are not loaded.

For example, suppose folder A contains folder B (empty), the file C.pcd, and folder D (containing 10 .pcd files). The result for each selection is as follows.

Select folder A → Only the single C.pcd file directly inside it is loaded.

Select folder B → Fails because there is no .pcd file.

Select the C.pcd file → The .pcd file is not loaded.

Select folder D → The 10 .pcd files inside folder D are loaded.

The AI-annotation setting configures the three options needed to enable the automatic AI-annotation feature.

Confidence: Sets a threshold. Predictions with a confidence below this threshold are excluded from the AI-annotation results. You can set a Confidence value for each Object Class.

Distance: A distance value that represents the AI-annotation results around the center point, forming a circle.

Track ID session: Sets whether to keep tracking the same object across frames.

What is a Track ID? A unique identifier assigned to each annotated object (box). The same object keeps the same Track ID even as frames change, so you can track it through multiple frames. Turning on Track ID session above enables this tracking, and a Track ID cannot be duplicated within a single frame.

X Factory recommends the optimal Confidence to maximize AI-annotation performance. Changing the set Confidence may affect the accuracy and quality of AI-annotation.

Confidence can only be set when creating a task. Once you run AI-annotation and a frame enters the In Progress state, you cannot change the Confidence, because the annotation has already been performed with the set Confidence value. To annotate with a different Confidence value, create a new task.

Excluding specific classes from detection

If there are classes you want to exclude from detection, you can exclude them in two ways.

When creating a project, uncheck the Object Classes you do not want to detect.

When creating a task within the project, uncheck the classes you do not want to detect.

The classes you uncheck are excluded from annotation.

(3) Performing annotation

There are three ways to perform annotation in X Factory.

1. AI-annotation

Per-frame AI-annotation

Select an individual frame and click the AI button at the top left of the screen to run AI-annotation on that frame. Per-frame AI-annotation can be run regardless of the frame's state. For accurate AI-annotation, select a model suitable for the task in Select Model. Confidence and Distance are set to the values configured when the project was created, and AI-Annotation Usage shows the number of frames that can be AI-annotated.

The available AI models and their environments are as follows.

AI model	Environment	Last updated
L_BiKal_mobility	Suitable for moving objects such as vehicles and robots	2025-05
L_BiKal_static_outdoor	Suitable for fixed outdoor infrastructure such as intersections and roads	2025-05

If edits are needed after AI-annotation, you can edit manually.

Full-frame AI-annotation

Click the AI Run button (Task > Run) to run AI-annotation on multiple frames at once. For accurate AI-annotation, select a model suitable for the task in Select Model. Confidence and Distance are set to the values configured when the project was created, and AI-Annotation Usage shows the number of frames that can be AI-annotated.

The available AI models and their environments are as follows.

AI model	Environment	Last updated
L_BiKal_mobility	Suitable for moving objects such as vehicles and robots	2025-05
L_BiKal_static_outdoor	Suitable for fixed outdoor infrastructure such as intersections and roads	2025-05

Full-frame AI-annotation and frame states

Full-frame AI-annotation (the AI Run button) applies only to frames in the Not Started state. If the task has Not Started frames, clicking AI Run will AI-annotate only the frames in the Not Started state.

Frames that have already been started (In Progress, Submitted, In Review, Completed — that is, any state other than Not Started) are excluded from full-frame AI-annotation. In that case, open each frame and run per-frame AI-annotation using the AI button at the top left of the screen.

Because multiple people often work on labeling together, this is designed so that work already in progress is not affected by a bulk run. To run full-frame AI-annotation on the entire task again, create a new task and run it while the frames are in the Not Started state.

AI Run is applied to the target frames at once and cannot be undone. We recommend testing AI-annotation on one or two frames first to confirm the settings before proceeding.

If edits are needed after AI-annotation, you can edit manually.

2. AI-assistance

AI-assistance helps annotation by generating boxes based on clustered points. After identifying the clustered points you want to annotate, enable AI-assistance to generate a 3D bounding box that fits the object. After creating the box, select the class you want and save it.

3. Manual Annotation

A feature that lets you annotate manually by hand.

Editing boxes

You can delete boxes that were detected incorrectly in the AI-annotation results, or add boxes directly to objects that were not detected.

Deleting a box

Deleting a single specific box

Click the box you want to delete and press the shortcut X to delete it.

Deleting boxes with the same Track ID across all frames

Press the View button (shortcut V) to switch to Top View.

Click the gear button at the bottom left and turn Object identification ON to display each object's Track ID above the box.

Click the box you want to delete, press the Inspector button at the top center, and click the Track ID copy button inside the Inspector.

Move to the frame that contains the box you want to delete. (Use the left/right arrow keys to move between frames.)

Click the Search Track ID button in the left toolbar.

Paste the copied Track ID into the input field at the top and press the shortcut X to delete the boxes with that Track ID.

Adding a box to an undetected object

You can manually label objects that were not detected.

When there is a box of the same size in the same frame (copy a box)

Click a box of the same size and press Ctrl + C to copy it.

Place the mouse at the center of the object you want to paste the box onto and press Ctrl + V to paste the box. The Track ID is not the same as the copied object; a new Track ID is assigned. (This is because Track IDs cannot be duplicated within a single frame.)

You can adjust the pasted box in the right panel.

When you want to create a box for an object with no fixed size (AI-assistance)

Click the AI-assistance feature in the left toolbar. (Shortcut Shift + S)

Drag over the object you want to create a box for and wait a moment.

A box is generated, and you can adjust its size and position in the right panel.

Copying and pasting a Track ID

You can apply one box's Track ID to a box in another frame.

Click the box that has the Track ID you want to copy, press the Inspector button at the top center, and click the Track ID copy button inside the Inspector.

Move to the frame that contains the box you want to apply the Track ID to. (Use the left/right arrow keys to move between frames.)

Click the box you want to apply it to, press the Inspector button at the top center, and click the check button under Paste Track ID inside the Inspector to apply it. (Track IDs cannot be duplicated within a single frame.)

(4) Creating annotation boxes

To create box data, the data must go through a Verify step. You can verify frame by frame, or verify all frames at once with the All Verify button. Once Verify is complete, the boxes can be downloaded.

Understanding work completion states

A frame's work state progresses in the order Not Started → In Progress → Submitted → In Review → Completed.

X Factory has two completion concepts: Task completion and AI-annotation completion. They are different concepts, so it helps to understand them separately.

Task completion

Task completion means that all frame work included in the task has been finished and reached the Completed state.

To complete a task's work, you must press the All Verify button. The All Verify button works only when there are frames in the In Review state.

When you press the All Verify button, only the frames in the In Review state change to the Completed state and the work is completed, and the All Verify button becomes disabled.

To complete the entire task, press the All Verify button after all frames are in the In Review state. It will then be shown as completed.

AI-annotation completion

This means that annotation has been completed for one or more frames within the task.

Exporting annotation results

Annotated data is saved in an exportable form and can be downloaded as a .box file. The downloaded data can be used for training or validation with other models.

Exporting is possible only when the task's work state is Completed.

Confirm that the task is in the Completed state.

In the Box Data area, click Schema, then click XFactorySchema.

After confirming that XFactorySchema has been added, click the Download button.

When you click Download, the file is downloaded.

When you unzip the downloaded file locally, you will see that it has been exported as .json files.

If the download in step 4 does not complete properly and you cannot find it locally, proceed with the following additional steps.

Click the Collection tab on the left to see the data in .box format.

Select the .box file and click the download button to download the data.

(5) Adding a class

To add a class, you need to configure the settings. You can add classes under Settings > Admin setting > Total classes. If you need to change the name of an existing class, you can rename that class. If you want to use a class extended from an existing class, you can add an extended class using the + button on the existing class. To use a class separate from existing classes, you can rename the Any Class or add an extended class to the Any Class.

Not satisfied with the results?

AI-annotation is a feature that automatically generates a draft of the results so you can start labeling quickly. Because the best settings differ depending on the sensor type, environment, and capture conditions of your data, the first run may produce results that differ from your expectations. This is a natural part of the process, and you can improve the results to fit your data using the methods below. You can adjust using the methods below, and if you still do not get the results you want, please contact us. Vueron's experts will help you.

1. Adjust the settings and run again

Adjust Confidence — Lowering the Confidence threshold detects more objects, while raising it keeps only the more certain objects. If many objects are missed, try lowering the value; if many are detected incorrectly, try raising it.

Change the model — In Select Model, choose a different model that fits your data and run it.

Adjust Distance — Adjust the Distance value that represents AI-annotation to fit your data.

Confidence and Distance are values set when creating a task, so to try different values, create a new task, adjust them, and run again. The model can be changed directly in Select Model at run time.

AI Run is applied to the entire task at once and cannot be undone. We recommend testing the settings on one or two frames first before applying to all. Also, running AI-annotation again deducts from your Free Trial run count.

2. Edit manually

You can edit the boxes created by AI. You can fix boxes whose position or size is off, change their class, or add missing objects. To quickly draw a box on clustered points, use AI-assistance; to draw from scratch, use Manual Annotation. For detailed methods such as deleting boxes, adding undetected objects, and managing Track IDs, see Editing boxes.

3. Check the data with PCD calibration

AI-annotation results are also affected by the calibration state of the point cloud data. If adjusting the settings and editing manually in X Factory still does not produce the results you want, check the calibration of your data.

With Vueron's PCD calibration product VueFlow, you can calibrate LiDAR point cloud data directly. When the data is correctly aligned, AI-annotation quality may improve.

For details, see the PCD Calibration Guide.

4. Get help from an expert

Still not getting the results you expect even after PCD calibration? Vueron's experts will produce results directly from your data. If you need help, contact xfactory@vueron.ai.

Modeling

Modeling and Deployment are AI training and deployment features, available only to customers using paid AI Annotation, through a separate sales inquiry. They are not provided in the Free Trial.

X Factory provides powerful pre-configured LiDAR perception AI models. Based on the Baseline models provided by X Factory, you can collect data and build a model optimized for your environment. This model is not one that runs on the cloud or a high-performance server, but an AI model optimized to run sufficiently on edge devices.

(1) AI model training

AI model training is performed by training a dataset created through Annotation, based on a baseline model. To train, select the data you want to use from the annotated datasets and separate it into a Training dataset and a Testing dataset. There are two methods for this.

Auto split: When you click Select data and choose a completed annotated dataset, the system automatically classifies it into a Training dataset and a Testing dataset.

User split: When you click Select data and choose a completed annotated dataset, you can choose the ratio of Training dataset to Testing dataset for each dataset yourself.

Then start training with the Start Training button.

The available AI models and their environments are as follows.

AI model	Environment	Last updated
Mobility_Baseline	Suitable for moving objects such as vehicles and robots	2025-05
Static-outdoor_Baseline	Suitable for fixed outdoor infrastructure such as intersections and roads	2025-05

(2) Analyzing model results

Once model training is complete, a model trained on the Baseline model is created. For the result analysis of this model, you can directly check quantitative evaluation results and perception results through the viewer, such as AP, Precision-Recall, and F1 Score, as well as the PR Curve, Recall/Confidence Score, and Precision/Confidence Score.

Deployment

Deployment is the task of deploying a trained model to an edge device. The models provided by X Factory are provided as an Engine file that is ready to run on edge devices. The types of edge devices currently supported are as follows.

Controller	Last updated
Embedded PC based on Nvidia AGX Orin	2025-05
Embedded PC based on Nvidia Orin NX	2025-05

Select the model you want to deploy from the trained models and click the Deployment button to start the deployment. Once deployment is complete, the model is created in Collection, and you can download it.

You can use the downloaded model on your desired edge device using the X Factory Inference API Manual.

Managing members

You can invite members to work together in your workspace and assign a role to each member.

Inviting members

Click Settings in the left menu.

Click the Invite members button at the top right of the Members tab.

Enter the email of the member you want to invite and click Invite. An invitation email is sent to the address you entered. You can invite several people at once.

Once an invited member joins the workspace through the invitation email, you can assign a role to that member. Until they join, they remain in Guest status and cannot be assigned a role. After joining, a member starts with the Labeler role, and an Admin changes it to the desired role.

Member invitations are available within your plan's member limit. If the limit is reached, remove a member to free up a seat and then invite again.

Roles and permissions

Each member has one of the following roles.

Role	What they can do
Owner	The person who created the workspace. Manages the plan and billing, and has all Admin permissions.
Admin	View and edit member information. Can invite members, change roles, and remove members.
Manager	Manages annotation work. Assigns frames to members and reviews the results to mark them complete.
Labeler	Annotates the frames assigned to them.

An Admin can change roles on the Members screen. When you select a member and change their role, access permissions are updated immediately. The workspace owner appears as Admin on the member screen.

Contact us

If you have any problems while using the product, contact xfactory@vueron.ai.

X Factory User Guide

Getting started with X Factory#

Start with the Free Trial#

Start by purchasing#

Understanding the Free Trial#

Usage limits#

Trial duration#

Collection (Data upload)#

Annotation#

(1) Create a project#

(2) Create a task#

(3) Performing annotation#

1. AI-annotation#

2. AI-assistance#

3. Manual Annotation#

Editing boxes#

Deleting a box#

Adding a box to an undetected object#

Copying and pasting a Track ID#

(4) Creating annotation boxes#

(5) Adding a class#

Not satisfied with the results?#

1. Adjust the settings and run again#

2. Edit manually#

3. Check the data with PCD calibration#

4. Get help from an expert#

Modeling#

(1) AI model training#

(2) Analyzing model results#

Deployment#

Managing members#

Inviting members#

Roles and permissions#

Contact us#

Getting started with X Factory

Start with the Free Trial

Start by purchasing

Understanding the Free Trial

Usage limits

Trial duration

Collection (Data upload)

Annotation

(1) Create a project

(2) Create a task

(3) Performing annotation

1. AI-annotation

2. AI-assistance

3. Manual Annotation

Editing boxes

Deleting a box

Adding a box to an undetected object

Copying and pasting a Track ID

(4) Creating annotation boxes

(5) Adding a class

Not satisfied with the results?

1. Adjust the settings and run again

2. Edit manually

3. Check the data with PCD calibration

4. Get help from an expert

Modeling

(1) AI model training

(2) Analyzing model results

Deployment

Managing members

Inviting members

Roles and permissions

Contact us