0

SynthCLIP: Are We Ready for a Fully Synthetic CLIP Training?

A framework called SynthCLIP uses synthetic text-image pairs generated by text-to-image networks and large language models to train CLIP models, achieving performance comparable to models trained on real data.

Year
2024
Venue
arXiv 2024
Authors
6
Hosting
Abstract onlyARXIV-DEFAULT

Cite

Notes

Only stored in your browser.

Attribution

Abstract & full text
arxiv.org/abs/2402.01832v2ARXIV-DEFAULT
TL;DR
Semantic Scholar
Attribution policy →

Abstract

We present SynthCLIP, a CLIP model trained on entirely synthetic text-image pairs. Leveraging recent text-to-image (TTI) networks and large language models (LLM), we generate synthetic datasets of images and corresponding captions at scale, with no human intervention. In this work, we provide an analysis on CLIP models trained on synthetic data. We provide insights on the data generation strategy, number of samples required, scaling trends, and resulting properties. We also introduce SynthCI-30M, a purely synthetic dataset comprising 30 million captioned images. Our code, trained models, and data, are released as open source at https://github.com/hammoudhasan/SynthCLIP

Authors

6