> For the complete documentation index, see [llms.txt](https://jtheta-ai.gitbook.io/docs.jtheta.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://jtheta-ai.gitbook.io/docs.jtheta.ai/domain-specific-workflow/lidar-image-annotation/export-lidar-datasets.md). # Export LiDAR Datasets Once annotation and validation are complete, export the finalized LiDAR dataset from **Datasets > \[Project Name]**. Before exporting, review the right-side configuration panel.

## **Pre-Export Validation Checklist** ### **1. Release Version** * Confirm the correct version (e.g., **v1.0**) is selected. * Click **Release New Version** if you need to freeze the current state. Each release version captures: * All annotated frames * Object instances and class mappings * Frame-level metadata * Media configuration > Only released versions can be exported. ### **2. Annotation Summary** Verify object counts (e.g., Pedestrian, Car, Cyclist). This ensures: * No missing annotations * No accidental deletions * Expected class distribution before export ### **3. Annotation Opacity (Validation Tool)** The **Annotation Opacity** slider adjusts bounding box transparency (visual only — does not affect export). Use it to: * Check 3D box alignment with the LiDAR point cloud * Verify box tightness and orientation * Inspect overlapping objects **Best practice:** * Lower opacity → inspect point cloud density * Higher opacity → validate box placement ### **4. Media Attributes** The **Media Attributes** section displays essential metadata associated with the selected LiDAR file. This information ensures data consistency, traceability, and export readiness. **Best practice:**\ Review and standardize all required attributes before export to ensure data quality and consistency. ## **Export Process** #### **1. Open Export Modal** Click **Download** (top-right). #### **2. Select Version** Example: **v1.0**\ Always confirm the intended release version. #### **3. Select Format (LiDAR Supported)** Available formats: * **KITTI** * **nuScenes** * **Supervisely** * **Custom JSON** * **Custom CSV** * **ROS Bag** ## Export Format Guide #### **(I) KITTI** Industry-standard format for: * 3D Object Detection * Autonomous Driving * Perception Model Training Includes annotations and calibration files compatible with KITTI-based pipelines. *** #### **(II) nuScenes** Designed for multi-sensor autonomous driving datasets. Supports: * LiDAR * Camera * Sensor calibration * Timestamp synchronization * Rich metadata *** #### **(III) Supervisely** Exports annotations in a structure compatible with the Supervisely platform for continued annotation or dataset management. *** #### **(IV) Custom JSON** Flexible structured format for custom machine learning pipelines. Ideal for: * Internal tools * Data processing * AI pipelines * Custom integrations *** #### **(V) Custom CSV** Exports annotation data in tabular format. Useful for: * Quality Assurance * Analytics * Dataset auditing * Reporting *** #### **(VI) ROS Bag** Exports the annotated dataset in **ROS Bag-compatible format**, enabling seamless integration with robotics development workflows. Ideal for: * Playback of recorded LiDAR sensor data * Testing and debugging perception algorithms * Robot simulation and validation * Offline development without requiring physical hardware * ROS and ROS 2 ecosystem integration ROS Bag export allows developers to replay synchronized sensor data together with annotations, making it easier to validate object detection, tracking, localization, and autonomous navigation pipelines. ## **Export Includes** Depending on format, the package may contain: * 3D bounding box parameters\ *(x, y, z, length, width, height, rotation/yaw)* * Class labels * Object instance IDs * Frame metadata * Sensor metadata * Multiview alignment data (if applicable) * Media attributes (format-dependent) * ROS Bag files (.bag or ROS 2 bag structure) containing synchronized LiDAR sensor recordings (for ROS Bag export) ## **Final Step** After confirming version and format: 👉 Click **Download** The system generates a structured LiDAR dataset package ready for training or deployment. Exports include: * 3D bounding box coordinates * Class labels * Frame metadata * Multiview alignment data (if applicable)