Stanford Cars: Vehicle Images with Robustness Testing Splits

A 2024-05-02 updated image dataset for vehicle recognition, uploaded to HuggingFace by tanganke. It contains 8,144 training images and 8,041 test images, with additional test splits for robustness evaluation. The splits include images with high contrast, Gaussian noise, impulse noise, JPEG compression, and motion blur.

Use Cases

• Train image classification models for vehicle make and model recognition based on the training split.
• Evaluate model robustness to image corruption based on the Gaussian noise and impulse noise test splits.
• Test model performance on compressed images based on the JPEG compression test split.
• Assess model resilience to motion blur based on the motion blur test split.
• Benchmark model generalization on high-contrast images based on the contrast test split.

Strengths

• Includes 8,144 images for training and 8,041 images for standard testing.
• Provides five distinct robustness test splits, each with 8,041 images, for evaluating model performance under corruption.
• Dataset was last updated on 2024-05-02, suggesting recent availability.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the full dataset is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

tanganke on HuggingFace, based on the Stanford Cars dataset.

Freshness

Last updated 2024-05-02 13:44:20; freshness should be verified.

Geography

Region 'us' indicated in platform tags.