English

Distilled Dual-Encoder Model for Vision-Language Understanding

Computation and Language 2022-10-18 v2 Computer Vision and Pattern Recognition

Abstract

We propose a cross-modal attention distillation framework to train a dual-encoder model for vision-language understanding tasks, such as visual reasoning and visual question answering. Dual-encoder models have a faster inference speed than fusion-encoder models and enable the pre-computation of images and text during inference. However, the shallow interaction module used in dual-encoder models is insufficient to handle complex vision-language understanding tasks. In order to learn deep interactions of images and text, we introduce cross-modal attention distillation, which uses the image-to-text and text-to-image attention distributions of a fusion-encoder model to guide the training of our dual-encoder model. In addition, we show that applying the cross-modal attention distillation for both pre-training and fine-tuning stages achieves further improvements. Experimental results demonstrate that the distilled dual-encoder model achieves competitive performance for visual reasoning, visual entailment and visual question answering tasks while enjoying a much faster inference speed than fusion-encoder models. Our code and models will be publicly available at https://github.com/kugwzk/Distilled-DualEncoder.

Keywords

Cite

@article{arxiv.2112.08723,
  title  = {Distilled Dual-Encoder Model for Vision-Language Understanding},
  author = {Zekun Wang and Wenhui Wang and Haichao Zhu and Ming Liu and Bing Qin and Furu Wei},
  journal= {arXiv preprint arXiv:2112.08723},
  year   = {2022}
}

Comments

EMNLP 2022

R2 v1 2026-06-24T08:19:57.843Z