ImageNet-1K Enriched with Captions, Bounding Boxes, and Label Issues

An enriched version of the ImageNet-1K dataset, last updated on 2024-09 16. It includes image captions, bounding boxes, and label issues, extending the dataset's utility beyond classification. The dataset was created by visual-layer and is hosted on Hugging Face.

Use Cases

• Image retrieval based on the provided image captions.
• Visual question answering using the combined image and caption data.
• Object detection model training or evaluation based on the bounding box annotations.
• Dataset curation and cleaning based on the identified label issues.

Strengths

• Enriched with three additional annotation types: captions, bounding boxes, and label issues.
• Based on the established ImageNet-1K benchmark dataset.
• Last updated on 2024-09-16, suggesting recent maintenance.

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.
• Row count, file formats, and exact size are unknown, which may limit suitability assessment.

Provenance

Source

visual-layer via Hugging Face.

Collection Method

An enrichment of the original ImageNet-1K dataset.

Time Range

Based on the original ImageNet collection period.

Freshness

Last updated 2024-09-16 11:13:12.

Geography

Global, based on the original ImageNet collection.