Aerial Scene Imagery With Multi-Label Annotations

AID_MultiLabel contains aerial scene images for multi-label classification tasks. The dataset is derived from the AID benchmark for aerial scene classification and includes multi-label annotations from a relation network study. Row and column counts are unspecified.

Use Cases

• Train a multi-label image classifier using aerial scene imagery and its associated category labels.
• Apply zero-shot image classification techniques to the aerial scene categories present in the dataset.
• Benchmark relation network or other multi-label models on the aerial imagery and its annotation labels.
• Analyze the distribution and co-occurrence of scene categories across the annotated image collection.

Strengths

• Dataset is in the public domain under a CC0 license, permitting unrestricted use.
• Includes annotations for multi-label classification, extending the original AID aerial scene benchmark.
• Available in Parquet format, compatible with libraries like Polars, Pandas, and Datasets.

Limitations

• Sample size is unspecified, making it difficult to assess scale for model training.
• The specific label categories, image count, and annotation schema are not detailed in the input.
• Last updated in April 2023, which may indicate staleness relative to current model architectures.

Provenance

Source

Hugging Face user jonathan-roberts1, derived from the AID benchmark dataset and a multi-label relation network study.

Collection Method

Curated from existing aerial imagery and multi-label annotations; specific gathering method unknown.

Time Range

null

Freshness

Last updated on 2023-04-03.

Geography

Region tag indicates 'US', but specific spatial coverage is unknown.