AutoMap Dataset

# AutoMap Dataset: High-Resolution Urban Mapping

The **AutoMap Dataset** is a high-fidelity multi-modal collection designed for urban mapping and autonomous navigation research. It features synchronized high-resolution imagery and dense point clouds from automotive-grade solid-state LiDARs, providing a robust benchmark for SLAM, 3D reconstruction, and sensor fusion.

## Table of Contents

1. [Dataset Overview](#1-dataset-overview)

2. [Hardware Setup & Mounting](#2-hardware-setup--mounting)

3. [Calibration & Synchronization](#3-calibration--synchronization)

4. [Collection Context](#4-collection-context)

5. [Directory Structure](#5-directory-structure)

6. [Coordinate Systems](#6-coordinate-systems)

7. [Usage & Getting Started](#7-usage--getting-started)

8. [Access & Licensing](#8-access--licensing)

## 1. Dataset Overview

* **Summary:** An urban dataset captured with high-resolution Basler cameras and dense Livox HAP solid-state LiDARs.

* **Total Duration:** 29 Minutes.

* **Environment:** Urban / Inner City (Hannover, Germany).

* **Weather:** Sunny.

* **Total Data Size:** 1.13 TB.

## 2. Hardware Setup & Mounting

The data was collected using a custom mobile mapping platform mounted on an Opel Vivaro van.

*CAD Mockup of the Sensors as Mounted on the Test Vehicle*

*Simulated FoV Coverage of the Mounted Sensors. Green: LiDAR Coverage. Purple: Camera Coverage*

*Front and Side View of the Mounted MSS During The Dataset Collection*

### Sensor Specifications

* **LiDAR**
* **Model:** 2 x [Livox HAP](https://www.livoxtech.com/de/hap) (Solid-State)
* **Configuration:** Left/Right Pair
* **Key Specs:** 120° FOV, Dense non-repetitive scanning
* **Camera**
* **Model:** 2 x [Basler a2A2840-67g5cBAS](https://www.baslerweb.com/de-de/shop/a2a2840-67g5cbas/)
* **Configuration:** Left/Right Pair
* **Key Specs:** 8.1 MP, 2840 x 2840 resolution
* **INS/GNSS**
* **Model:** [OxTS AV200](https://www.oxts.com/solutions/inertial-navigation-solutions/navigation-hardware/av200/)
* **Configuration:** Roof-mounted
* **Key Specs:** Dual Antenna, RTK/PPK capable

### Mounting & Coverage

The sensors are mounted in a rigid frame atop the Opel Vivaro, optimized for wide area coverage, and minimum overlap between the solid-state LiDAR FOVs and the high-resolution cameras.

## 3. Calibration & Synchronization

Precision is maintained through rigorous calibration routines and sub-millisecond hardware synchronization.

### Calibration Procedure

* **Intrinsic (Camera):** Checkerboard-based calibration using the ROS Calibration Toolbox. Parameters are stored in `calibration.json`.

* **Intrinsic (IMU):** Allan Variance analysis via the ROS Toolbox.

* **Extrinsic (Sensor-to-Sensor):**

* *Camera-LiDAR:* Initial extrinsic parameters for left and right pairs were estimated using ACSC toolbox.

* *Camera-IMU:* The extrinsic transformation between the Left Camera and the INS IMU was estimated using the Kalibr toolbox.

* *Global Alignment:* The two sensor pairs (LiDAR-Camera pairs) were ultimately calibrated and aligned using the cameras with Ground Control Points (GCPs) in a known external reference frame to ensure absolute spatial accuracy.

* *Transformations: All final extrinsic matrices (`T_sensor_to_left_lidar`) derived from this multi-step process are provided in `calibration.json`.

### Synchronization & Status

* **Grandmaster Clock:** All hardware is synchronized to GPS time.

* **Protocols:**

* *PTP:* Used for synchronization between INS and Camera.

* *gPTP:* Used for LiDAR synchronization.

* **Data Status:**

* *Images:* Provided as Raw (Bayer RG8) lossless `.png` format .

* *LiDAR:* Provided as raw and undistorted `.pcd` (corrected via internal LiDAR IMUs).

* *Navigation:* INS data is PPK-processed at 100 Hz, including GNSS and raw IMU logs in a `.csv` format.

## 4. Collection Context


*Route of the Vehicle During the Dataset Collection*

**Location:** Hannover, Germany

The dataset was recorded in the inner city of Hannover. The trajectory covers a mix of narrow and wide European streets.

### Conditions

* **Time of Day:** Afternoon.

* **Traffic:** Variable density featuring multiple "stop-and-go" scenarios, typical for mid-day urban environments.

* **Complexity:** Includes interactions with pedestrians, cyclists, and trams.

## 5. Directory Structure


## 6. Coordinate Systems

Local Sensor Frames: Each sensor's data (images, point clouds, IMU) is provided in its respective local coordinate frame. To transform data between sensors, refer to the transformation matrices in calibration.json.

* **Reference Frame:** The Left LiDAR serves as the primary coordinate reference for the sensor rig.

* *Global References:*

* Lateral: ETRS89 / UTM zone 32N (EPSG:25832).

* Vertical: DHHN2016 (German Main Height Network / EPSG:7837).

## 7. Usage & Getting Started

To Be Released

## 8. Licensing

### License

This dataset is licensed under a Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) License. You are free to share and adapt the material for academic and research purposes, as long as appropriate credit is given and the dataset is not used for commercial purposes.