Russell Howes

2 accepted papers

2024

Demystifying CLIP Data

Hu Xu, Saining Xie, Xiaoqing Tan, Po-Yao Huang, Russell Howes, Vasu Sharma +4

ICLR 2024spotlight

Contrastive Language-Image Pre-training (CLIP) is an approach that has advanced research and applications in computer vision, fueling modern recognition systems and generative models. We believe that the main ingredient to the success of CLIP is its \textit{data} and \textit{not} the \textit{model}…

Cited by 166 Source ↗PDF ↗code ↗Scholar ↗

2023

CiT: Curation in Training for Effective Vision-Language Data

Hu Xu, Saining Xie, Po-Yao Huang, Licheng Yu, Russell Howes, Gargi Ghosh +2

ICCV 2023poster

Large vision-language models are generally applicable to many downstream tasks, but come at an exorbitant training cost that only large institutions can afford. This paper trades generality for efficiency and presents Curation in Training (CiT), a simple and efficient vision-text learning algorithm…

Cited by 28 PDF ↗code ↗Scholar ↗