AltCLIP: Altering the Language Encoder in CLIP for Extended Language Capabilities
This work addresses the need for multilingual multimodal models, but it is incremental as it builds directly on CLIP with a modified encoder.
The authors tackled the problem of extending CLIP's language capabilities to bilingual/multilingual settings by replacing its text encoder with XLM-R and using a two-stage training approach, achieving new state-of-the-art performances on tasks like ImageNet-CN and COCO-CN while maintaining performance close to CLIP on most tasks.
In this work, we present a conceptually simple and effective method to train a strong bilingual/multilingual multimodal representation model. Starting from the pre-trained multimodal representation model CLIP released by OpenAI, we altered its text encoder with a pre-trained multilingual text encoder XLM-R, and aligned both languages and image representations by a two-stage training schema consisting of teacher learning and contrastive learning. We validate our method through evaluations of a wide range of tasks. We set new state-of-the-art performances on a bunch of tasks including ImageNet-CN, Flicker30k-CN, COCO-CN and XTD. Further, we obtain very close performances with CLIP on almost all tasks, suggesting that one can simply alter the text encoder in CLIP for extended capabilities such as multilingual understanding. Our models and code are available at https://github.com/FlagAI-Open/FlagAI.