Published July 30, 2024
| Version 1.0
Dataset
Open
CLIP-Embedded RedCaps Text-Image Dataset
Creators
-
1.
Massachusetts Institute of Technology
Description
This dataset was created by applying the CLIP embedding to the RedCaps dataset. Queries are generated by OpenAI's GPT model simulating textual queries searching multimodal content, embedded via CLIP. The data was curated by Desai, Kaul, Aysola, and Johnson from data collected by Reddit, and further curated into vector data by Engels for this work.
Usage of the dataset itself is subject to Reddit terms, Reddit User Agreeement, Content Policy, and Privacy Policy (quoted from Desai et. al.'s accompanying paper for the image-and-text dataset). Usage of the queries are subject to OpenAI terms. Among others, OpenAI terms prohibits using the query component of this dataset to develop models that compete against OpenAI.
Files
Files
(23.8 GB)
| Name | Size | Download all |
|---|---|---|
|
md5:a6221cc0a4103af7e0f06f87bd989a0a
|
23.8 GB | Download |
Additional details
Identifiers
- arXiv
- arXiv:2402.00943
Related works
- Is published in
- arXiv:2111.11431 (arXiv)
- Is supplemented by
- arXiv:2402.00943 (arXiv)
Dates
- Available
-
2024-07-30
References
- embedded-redcaps