Conceptual Captions 12 Million Image-Text Pairs

Conceptual Captions 12M (CC12M) contains 12 million image-text pairs designed for vision-and-language pre-training. It was created by pixparse and is a relaxed version of the CC3M dataset pipeline. The dataset instance was last updated on Hugging Face in December 2023.

Use Cases

• Train image captioning models using the paired image and text fields.
• Perform vision-language pre-training for tasks like visual question answering using the multimodal pairs.
• Benchmark model performance on large-scale image-text alignment using the 12 million pairs.
• Fine-tune models for zero-shot image classification using the descriptive text labels.

Strengths

• Contains 12 million image-text pairs.
• Built as a relaxed version of the established CC3M pipeline.

Limitations

• Specific column names, file formats, and data size are unknown.
• The temporal and geographic coverage of the images and text are unknown.

Provenance

Source

pixparse on Hugging Face.

Collection Method

Data collection pipeline is a relaxed version of the one used for Conceptual Captions 3M (CC3M).

Time Range

null

Freshness

null

Geography

null