Summarized short captions of ShareGPT4v for training

beichenzbc / Long-CLIP

[ECCV 2024] official code for "Long-CLIP: Unlocking the Long-Text Capability of CLIP"

Apache License 2.0

690 stars 33 forks source link

Summarized short captions of ShareGPT4v for training #59

Closed ivonajdenkoska closed 3 months ago

ivonajdenkoska commented 4 months ago

Hi, thanks for the great work!

Can you explain how you obtained the summarized short captions of ShareGPT4v for training?

The paper mentions the usage of these short captions, but we cannot find them in the repo or in HugginngFace (https://huggingface.co/datasets/Lin-Chen/ShareGPT4V). Can you release them? Thanks a lot in advance!

beichenzbc commented 4 months ago

Thanks, that's a good question. ShareGPT4v only contains long captions. However, the first sentence is usually a short summary sentence like 'the image showcases ......'. Therefore, we can take the first sentence as the short caption.