1,000,000+ annotations for vehicle detection and orientation classification. This dataset enables the training of standard object detection networks to predict both bounding boxes and vehicle headings.
Use Cases
- Train a multi-task object detection model to predict vehicle bounding boxes and orientation classes
- Evaluate the performance of orientation classification heads on a large-scale set of over one million samples
- Develop autonomous driving perception modules that require precise vehicle heading information
Strengths
- Contains over 1,000,000 individual vehicle annotations
- Includes labels for both object detection (bounding boxes) and orientation classification
- Designed for use with standard object detection network architectures