Distilled Dual-Encoder Model for Vision-Language Understanding
This work addresses the trade-off between accuracy and speed in vision-language models for applications requiring efficient real-time processing, though it is incremental as it builds on existing dual-encoder and distillation methods.
The paper tackles the problem of dual-encoder models having insufficient interaction for complex vision-language tasks by proposing cross-modal attention distillation from a fusion-encoder model, achieving competitive performance on tasks like visual reasoning and visual question answering with much faster inference speed.
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.