Chinese CLIP: Contrastive Vision-Language Pretraining in Chinese

TL;DR

This paper introduces Chinese CLIP, a set of large-scale Chinese vision-language models trained on a new dataset, employing a two-stage pretraining method, achieving state-of-the-art results on multiple benchmarks.

Contribution

It constructs a large-scale Chinese image-text dataset and develops multiple Chinese CLIP models with a novel two-stage pretraining approach.

Findings

Chinese CLIP achieves state-of-the-art results on MUGE, Flickr30K-CN, and COCO-CN.

Models perform competitively in zero-shot image classification on ELEVATER.

Two-stage pretraining enhances model performance.

Abstract

The tremendous success of CLIP (Radford et al., 2021) has promoted the research and application of contrastive learning for vision-language pretraining. In this work, we construct a large-scale dataset of image-text pairs in Chinese, where most data are retrieved from publicly available datasets, and we pretrain Chinese CLIP models on the new dataset. We develop 5 Chinese CLIP models of multiple sizes, spanning from 77 to 958 million parameters. Furthermore, we propose a two-stage pretraining method, where the model is first trained with the image encoder frozen and then trained with all parameters being optimized, to achieve enhanced model performance. Our comprehensive experiments demonstrate that Chinese CLIP can achieve the state-of-the-art performance on MUGE, Flickr30K-CN, and COCO-CN in the setups of zero-shot learning and finetuning, and it is able to achieve competitive performance in zero-shot image classification based on the evaluation on the ELEVATER benchmark (Li et al., 2022). We have released our codes, models, and demos in https://github.com/OFA-Sys/Chinese-CLIP