Dataset AIL Albus

This page contains data recordings of a part of the research project for the Ministry of Transport of the Slovak Republic, which was testing the readiness of Slovak road infrastructure for the deployment of intelligent transportation. To create this dataset, 96 individual routes were conducted on various types of Slovak roads. The roads were selected based on their type or environment, which was the following:

• City – Bratislava – ±13km long route across urban roads in Bratislava,
• Highway – D1 – ±12km long route on a 3-lane highway,
• Highway – D2 – ±32km long route on a 2-lane highway that contains tunnel Sitina with a length of 1415m,
• 1st class – I/61 – ±30km long route on the 1st class road in Bratislava region that contains rural and urban environment,
• 2nd class – II/503 – ±25km long route on a 2nd class road that contains hills and forest,
• 3rd class – various – ±19km long route across several 3rd class roads that contains local rural transportation with limited traffic.

For each type of road, we provide recordings of different weather and light conditions. We used the following methodology when sorting the samples into a specific category:

• Direct sunlight – No atmospheric precipitation, the sky was clear or with low clouds, the sunlight was direct, and objects in the camera’s field of view cast clearly recognizable sharp shadows,
• Normal daylight – No atmospheric precipitation, the sky was completely overcast or with very high cloud cover, sunlight was scattered and objects in the camera’s field of view cast minimal or no shadows,
• Rain or fog – Atmospheric precipitation (rain or sleet) or fog was present, resulting in a wet road surface and water condensation on the front camera lens,
• Dawn or twilight – No atmospheric precipitation, cloud cover was not decisive. The drive was carried out between the beginning of daylight and sunrise above the horizon, or between sunset below the horizon and the time of complete darkness.

Usage policy

All datasets and benchmarks on this page are copyright by us and published under the Creative Commons Attribution 4.0 (CC BY 4.0) license. This means that you must attribute the work in the manner specified by the authors, you may not use this work for commercial purposes and if you alter, transform, or build upon this work, you may distribute the resulting work only under the same license.

When using this dataset, please cite following research paper: [TO BE ADDED]

Sensor Setup

On this section, we provide information regarding the recording platform and sensor configuration utilized for capturing this dataset. For all drives, Land Rover Defender equipped with an AIL System Solution rooftop box, which is a custom-built proprietary solution of the Automotive Innovation Lab research laboratory at STU. We used following sensors in rooftop box for data recording:

Front camera

For recording of camera frames we used D3RCM-OV10640-953 rugged camera module by DesignCore^TM with 1.3 MegaPixel and FPD-LinkTM III interface. This camera suits automotive and industrial production needs. The lens Sunny 4083 provides 52 degree horizontal field of view (FOV) with aperture of 1.6 and 6.14 mm focal length. The resolution of each frame is 1280 × 1080, meaning the frames are in 4:3 ratio.

GNSS Sensor

The GNSS development kit from the company U-blox C102 – F9R is a device for determining high-precision position. The module has the option of connecting a dual-band external antenna, which together with Real-Time Kinematics (RTK) over the Internet Protocol using NTRIPv3 can ensure the fixed accuracy of the position.

Mobile network modem

A network modem and router, Teltonika RUTX50, that provides connection to 2G, 3G, 4G (LTE) and 5G NSA mobile network. The device has a modem that supports all transmission bands of mobile operators used in the territory of the Slovak Republic.

Data Description

We offer four recordings in ZIP format for each combination of road type and weather or light conditions. Each recorded route includes the following data:

• Processed (anonymized) color video images (stored in JPG format)
• Raw data from GNSS and Mobile network modem (latitude, longitude, height above mean sea level, ggs, gc, horizontal accuracy, satellite, ping, download speed, upload speed, internet mode, stored in CSV format)
• Road lane detection data (name of the video frame, number of detecting lanes, points of the detected lane, length of the detected lane, stored in JSON and CSV format)
• Roadsigns detection data (name of video frame, category of roadsign, probability score, x and y coordinates of detected roadsign, stored in JSON and CSV format)

Examples

Here, we provide illustrative examples of video frames from various datasets, each representing different road types and environmental conditions.

Privacy statement

Important: The recorded videos are anonymized using a program that has obfuscated license plates and individuals on the roads. However, it’s possible that not all videos have been anonymized, and some video frames might still raise privacy concerns. If you have any issues regarding non-anonymized video frames or potential privacy breaches, please don’t hesitate to contact our team.

In case of interest, it is possible to make available the non-anonymized and non-preprocessed version of the dataset by mutual agreement.

Downloadable Datasets