COCO Image Captions with Karpathy Split

A version of the COCO dataset prepared for image captioning tasks using the Karpathy split. The dataset was created by the author yerevann and was last updated on Hugging Face in October 2022.

Use Cases

• Train image captioning models using image and caption pairs.
• Evaluate caption generation performance on a standard split.
• Benchmark vision-language models on the COCO dataset.
• Develop object recognition systems using annotated images.

Strengths

• Based on the widely-used COCO dataset.
• Uses a standard split for consistent evaluation.

Limitations

• Specific row count, column details, and file formats are unknown.
• Dataset size and sample data are unavailable for assessment.

Provenance

Source

COCO dataset.

Collection Method

Karpathy split applied to COCO.

Freshness

Last updated on Hugging Face in 2022.