SUN397: 397 Scene Categories for Image Recognition

108,754 images across 397 scene categories, with at least 100 images per category. The dataset is a subset of the SUN database, created for research purposes by tanganke and last updated on Hugging Face in May 2024.

Use Cases

• Train scene classification models based on the 397 labeled categories.
• Benchmark image recognition algorithms on a diverse set of indoor and outdoor scenes.
• Perform visual dataset analysis on the distribution of images across categories.

Strengths

• Contains 108,754 images in total.
• Covers 397 distinct scene categories.
• Provides a minimum of 100 images per category for balanced representation.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.

Provenance

Source

Subset of the SUN dataset.

Freshness

Last updated 2024-05-16 02:41:24; freshness should be verified.