License Plate Detection Dataset with Train and Validation Splits

8,823 images are provided for training, validation, and testing computer vision models. The dataset was created by keremberke and published on HuggingFace in January 2023. It contains bounding box annotations for license plate objects.

Use Cases

• Train an object detection model to locate 'license_plate' bounding boxes in vehicle images.
• Benchmark model performance across the 'train', 'valid', and 'test' splits containing 6176, 1765, and 882 images respectively.
• Evaluate model accuracy on the 'license_plate' label using the pre-defined test set.
• Fine-tune a pre-trained detector using the 'train' split annotations.

Strengths

• Provides 8,823 annotated images for object detection tasks.
• Includes pre-split subsets: 6176 training, 1765 validation, and 882 test images.

Limitations

• No information on geographic origin or diversity of license plate styles.
• Unknown image resolution, quality, and potential class imbalance.

Provenance

Source

keremberke on HuggingFace, originally sourced from Roboflow.

Collection Method

null

Time Range

null

Freshness

Last updated on HuggingFace in January 2023.

Geography

null