Heb Clip: Hebrew Image Captions for Vision-Language Models

A collection of 4 million Hebrew image captions derived from the Recap-DataComp-1B dataset, published by NVIDIA on August 27, 2024. It is designed for training vision-language models like CLIP for the Hebrew language, providing text captions paired with references to pre-computed image embeddings.

Use Cases

• Training Hebrew-language CLIP models based on the provided image caption pairs.
• Fine-tuning vision-language models for Hebrew image-text retrieval tasks.
• Benchmarking cross-lingual vision-language alignment performance.
• Augmenting existing multimodal datasets with Hebrew-language annotations.

Strengths

• 4 million captions provide a substantial text corpus for training.
• Captions are paired with references to pre-computed image embeddings, facilitating model training.

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 for the full dataset is unknown, which may limit suitability assessment.

Provenance

Source

NVIDIA

Collection Method

Translated from the Recap-DataComp-1B dataset.

Freshness

Last updated 2024-08-27 11:44:44; freshness should be verified.