The inD Dataset: A Drone Dataset of Naturalistic Road User Trajectories at German Intersections (1911.07602v1)

Abstract: Automated vehicles rely heavily on data-driven methods, especially for complex urban environments. Large datasets of real world measurement data in the form of road user trajectories are crucial for several tasks like road user prediction models or scenario-based safety validation. So far, though, this demand is unmet as no public dataset of urban road user trajectories is available in an appropriate size, quality and variety. By contrast, the highway drone dataset (highD) has recently shown that drones are an efficient method for acquiring naturalistic road user trajectories. Compared to driving studies or ground-level infrastructure sensors, one major advantage of using a drone is the possibility to record naturalistic behavior, as road users do not notice measurements taking place. Due to the ideal viewing angle, an entire intersection scenario can be measured with significantly less occlusion than with sensors at ground level. Both the class and the trajectory of each road user can be extracted from the video recordings with high precision using state-of-the-art deep neural networks. Therefore, we propose the creation of a comprehensive, large-scale urban intersection dataset with naturalistic road user behavior using camera-equipped drones as successor of the highD dataset. The resulting dataset contains more than 11500 road users including vehicles, bicyclists and pedestrians at intersections in Germany and is called inD. The dataset consists of 10 hours of measurement data from four intersections and is available online for non-commercial research at: http://www.inD-dataset.com

• The paper introduces a comprehensive dataset capturing over 11,500 naturalistic road user trajectories using drones, advancing automated driving research.
• The methodology employs semantic segmentation and deep neural networks to accurately track vehicles, pedestrians, and bicyclists at urban intersections.
• The dataset fills a gap in scale and diversity, providing a robust foundation for safety validation, traffic simulation, and improved road user prediction models.

Analysis of the inD Dataset: A Large-Scale Collection of Naturalistic Road User Trajectories

The publication of the inD dataset represents a significant contribution to the field of automated driving research. The authors have addressed an unmet demand for comprehensive datasets that can be leveraged to improve road user prediction models, scenario-based safety validation, and other facets critical to automated vehicle technologies. Utilizing drones as a methodological paradigm for data collection, this dataset fills a crucial gap in existing literature by focusing specifically on urban intersections, environments characterized by complex and diverse scenarios.

Overview of Dataset Methodology

The inD dataset comprises more than 11,500 road users captured over 10 hours from four urban intersections in Germany. This extensive dataset includes vehicles, pedestrians, and bicyclists, offering a diverse array of road user interactions. The methodology employed a camera-equipped drone flying at altitudes up to 100 meters, allowing unobtrusive data collection from a bird’s-eye perspective. This approach mitigates issues such as occlusions and behavioral distortions commonly associated with ground-level data collection methods.

The authors have applied state-of-the-art deep learning techniques to ensure high accuracy in detection, classification, and tracking of road users. By employing semantic segmentation through deep neural networks, the dataset provides precise trajectory information with a typical positioning error at a pixel level. This accuracy is critical for developing reliable automated driving algorithms that can effectively interpret traffic at intersections.

Comparison to Existing Datasets

The authors conducted a thorough review of existing datasets and found that many have limitations either in scale, diversity, or representativeness of data. For instance, while the Stanford Drone Dataset also employs drones for data collection, the majority of its trajectories are for pedestrians, limiting its applicability in vehicle-centric automated driving scenarios. In contrast, the inD dataset offers a well-balanced distribution of road user types on public German intersections, which are imperative for automated driving applications.

Implications and Future Directions

The availability of the inD dataset supports multiple research avenues, including safety validation for highly automated driving systems, traffic simulation models, traffic analysis, and road user prediction models. By providing this comprehensive dataset, the authors enable the development of more sophisticated, reliable algorithms capable of navigating complex urban traffic environments.

Looking forward, the practical implications of this dataset in AI include advancing machine learning models that can accurately anticipate and respond to the intricate interactions encountered at intersections. Theoretically, this dataset could also facilitate research into new paradigms of unsupervised or semi-supervised learning, where models are trained on the extensive, unlabeled data that the inD dataset provides.

In conclusion, the inD dataset enhances the resources available for research in automated driving. By addressing the challenges of data scarcity in complex urban settings, it lays the groundwork for future advancements in intelligent vehicle technologies. Continued exploration and utilization of such datasets ensure the progression of automated driving systems from theoretical constructs to real-world deployment.