Zenseact Edge AnnotatationZ Dataset
This dataset is provided by Zenseact with the intent of bringing real-world challenges to academia and startups. It consists of 6,666 image sequences captured by Zenseact developmental vehicles and was collected from highway, country, and urban roads in and around Warsaw, Poland. This dataset refers to the dataset used in the Edge AnnotationZ Challenge. For the large-scale version of Zenseact Open Dataset, please visit zod.zenseact.com.
In short
Content
6,666 three-frame image sequences of highway, country, and urban roads. Day and night, varying weather conditions.
Author
The dataset is led by Mina Alibeigi, Ph.D. in Machine Intelligence & Robotics. It is made possible by Daria Motorniuk, Oleksandr Panasenko, Jakub Bochynski, Dónal Scanlan, and Benny Nilsson. Special thanks to Jenny Widahl, Jonas Ekmark, Bolin Shao, and Erik Rosén for their support, comments, and suggestions.
Data Type
The dataset is real-world data and includes real-world traffic scenarios from highways, country roads, and urban areas in and around Warsaw, Poland, recorded during a three-week timespan. The data was collected during the day and night under varying weather conditions.
Anonymization
The dataset was anonymized. Faces and vehicle license plates are anonymized using the BrighterAI tools for GDPR compliance and the intent to preserve every identity on roads. There are two anonymized RGB images provided per frame: One with blurred faces and license plates and the other with faces and license plates replaced by Deep Natural Anonymization. The technique is based on generative AI and leads to a minimal pixel impact. Information like the line of sight of pedestrians is maintained and the anonymization method supports machine learning use cases in the best possible way.
Annotations
Core frames in the sequences are annotated with the GeoJSON format.
Size
3,4TB
Access
The dataset is available for all AI Sweden partners.
Terms and conditions
To use this dataset, you must comply with the Zenseact Open Dataset Terms and Conditions available below.
The Zenseact Edge AnnotatationZ Dataset (part of Zenseact Open Dataset) is the property of Zeanseact AB and it is licensed under CC BY - SA 4.0. Please check out this explanatory file for a better understanding of what the Creative Commons license implies.
Dataset specifics
This first release of the dataset includes real-world traffic scenarios from highways, country roads, and urban areas in and around Warsaw, Poland, recorded during a three-week timespan. The data was collected during day and night under varying weather conditions.
The dataset is a multimodal dataset containing 6,666 image sequences in total. Each sequence is composed of three consecutive high-resolution (8MP) RGB camera images with 30 Hz frequency in addition to their corresponding LiDAR point clouds, high-precision GPS measurements (a.k.a. OXTS), and comprehensive vehicle bus data (e.g., IMU, lateral and longitudinal velocity and acceleration, steering wheel angle, brake pedal pressure).
Faces and vehicle license plates are anonymized using the brighterAI tools for GDPR compliance and the intent to preserve every identity on roads. There are two anonymized RGB images provided per frame: One with blurred faces and license plates and the other with faces and license plates replaced by Deep Natural Anonymization. The technique is based on generative AI and leads to a minimal pixel impact. Information like line of sight of pedestrians is maintained and the anonymization method supports machine learning use cases in the best possible way.
The Zenseact Edge AnnotatationZ Dataset also includes fine annotations for various tasks, including instance and semantic segmentation annotations for the road and 2D and 3D bounding boxes for the static and dynamic objects. The middle frame, also called the core frame, in each 3-frame sequence is annotated for several annotation tasks, allowing multi-task learning.
More information about the dataset.
Use cases to date
1. Edge AnnotationZ Challenge
This dataset was available in the Data Factory for the participating teams during the Edge AnnotationZ Challenge.
Further reading
→ Edge AnnotationZ Challenge
Access
The dataset is available for all AI Sweden partners. Contact Beatrice Comoli for further instructions on how to access the data. If you are interested in becoming a partner of AI Sweden, getting access to the partner benefits, including the Data Factory and datasets, or in sharing a dataset or a model, please feel free to reach out.
