Saved in:
Bibliographic Details
Main Authors: Lei, Youbo, He, Feifei, Chen, Chen, Mo, Yingbin, Li, Si Jia, Xie, Defeng, Lu, Haonan
Format: Preprint
Published: 2023
Subjects:
Online Access:https://arxiv.org/abs/2310.19654
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914738058821632
author Lei, Youbo
He, Feifei
Chen, Chen
Mo, Yingbin
Li, Si Jia
Xie, Defeng
Lu, Haonan
author_facet Lei, Youbo
He, Feifei
Chen, Chen
Mo, Yingbin
Li, Si Jia
Xie, Defeng
Lu, Haonan
contents Due to the success of large-scale visual-language pretraining (VLP) models and the widespread use of image-text retrieval in industry areas, it is now critically necessary to reduce the model size and streamline their mobile-device deployment. Single- and dual-stream model structures are commonly used in image-text retrieval with the goal of closing the semantic gap between textual and visual modalities. While single-stream models use deep feature fusion to achieve more accurate cross-model alignment, dual-stream models are better at offline indexing and fast inference.We propose a Multi-teacher Cross-modality Alignment Distillation (MCAD) technique to integrate the advantages of single- and dual-stream models. By incorporating the fused single-stream features into the image and text features of the dual-stream model, we formulate new modified teacher similarity distributions and features. Then, we conduct both distribution and feature distillation to boost the capability of the student dual-stream model, achieving high retrieval performance without increasing inference complexity.Extensive experiments demonstrate the remarkable performance and high efficiency of MCAD on image-text retrieval tasks. Furthermore, we implement a lightweight CLIP model on Snapdragon/Dimensity chips with only $\sim$100M running memory and $\sim$8.0ms search latency, achieving the mobile-device application of VLP models.
format Preprint
id arxiv_https___arxiv_org_abs_2310_19654
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle MCAD: Multi-teacher Cross-modal Alignment Distillation for efficient image-text retrieval
Lei, Youbo
He, Feifei
Chen, Chen
Mo, Yingbin
Li, Si Jia
Xie, Defeng
Lu, Haonan
Computer Vision and Pattern Recognition
Artificial Intelligence
Due to the success of large-scale visual-language pretraining (VLP) models and the widespread use of image-text retrieval in industry areas, it is now critically necessary to reduce the model size and streamline their mobile-device deployment. Single- and dual-stream model structures are commonly used in image-text retrieval with the goal of closing the semantic gap between textual and visual modalities. While single-stream models use deep feature fusion to achieve more accurate cross-model alignment, dual-stream models are better at offline indexing and fast inference.We propose a Multi-teacher Cross-modality Alignment Distillation (MCAD) technique to integrate the advantages of single- and dual-stream models. By incorporating the fused single-stream features into the image and text features of the dual-stream model, we formulate new modified teacher similarity distributions and features. Then, we conduct both distribution and feature distillation to boost the capability of the student dual-stream model, achieving high retrieval performance without increasing inference complexity.Extensive experiments demonstrate the remarkable performance and high efficiency of MCAD on image-text retrieval tasks. Furthermore, we implement a lightweight CLIP model on Snapdragon/Dimensity chips with only $\sim$100M running memory and $\sim$8.0ms search latency, achieving the mobile-device application of VLP models.
title MCAD: Multi-teacher Cross-modal Alignment Distillation for efficient image-text retrieval
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2310.19654