Alapfogalmak
IDD-X is a large-scale dual-view driving video dataset that provides comprehensive ego-relative annotations for multiple important road objects and their corresponding explanations in dense and unstructured traffic environments.
Kivonat
The IDD-X dataset is designed to address the challenge of understanding the influence of road and traffic conditions on an ego vehicle's driving behavior, particularly in complex traffic situations found in developing countries. Unlike existing datasets that focus on structured and sparse traffic scenarios, IDD-X captures dense, heterogeneous, and unstructured traffic environments with multiple important road objects simultaneously affecting the ego vehicle's driving decisions.
The key highlights of the IDD-X dataset include:
- Dual-view driving videos (front and rear) with 697K bounding boxes, 9K important object tracks, and 1-12 objects per video.
- Comprehensive ego-relative annotations for 10 categories of important road objects and 19 explanation label categories, covering a diverse range of complex interaction patterns between the ego vehicle and surrounding entities.
- The first dataset to consider rearview information for important object annotations, providing a more complete representation of the driving environment.
- Custom-designed deep network architectures for multiple important object localization and per-object explanation prediction, serving as foundational components for understanding the nuanced relationships between road conditions and ego vehicle's driving behavior.
The IDD-X dataset and the introduced prediction models form a comprehensive framework for studying how road conditions and surrounding entities affect driving behavior in complex traffic situations, particularly in developing countries.
Statisztikák
The dataset contains 697K bounding boxes for important road objects.
There are 9K important object tracks in the dataset.
The number of important objects per driving scenario ranges from 1 to 12.
Idézetek
"Understanding the influence of road and traffic conditions on ego vehicle's driving behavior is crucial for enabling explainability in automated driving decision-making."
"Unlike Western countries, developing nations contain dense and unstructured traffic situations, with heterogeneous road occupants (two-wheelers, animals, three-wheelers, etc), and static road objects (speed breakers, potholes, and traffic lights, etc.)."
"By encompassing both front and rear views, IDD-X enables a more comprehensive analysis of driving behavior, providing a panoramic view of objects, their interactions, and the intricate cues that influence the driver's choices."