We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.CV

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computer Vision and Pattern Recognition

Title: Bridge-Tower: Building Bridges Between Encoders in Vision-Language Representation Learning

Abstract: Vision-Language (VL) models with the Two-Tower architecture have dominated visual-language representation learning in recent years. Current VL models either use lightweight uni-modal encoders and learn to extract, align and fuse both modalities simultaneously in a cross-modal encoder, or feed the last-layer uni-modal features directly into the top cross-modal encoder, ignoring the semantic information at the different levels in the deep uni-modal encoders. Both approaches possibly restrict vision-language representation learning and limit model performance. In this paper, we introduce multiple bridge layers that build a connection between the top layers of uni-modal encoders and each layer of the cross-modal encoder. This enables comprehensive bottom-up interactions between visual and textual representations at different semantic levels, resulting in more effective cross-modal alignment and fusion. Our proposed Bridge-Tower, pre-trained with only $4$M images, achieves state-of-the-art performance on various downstream vision-language tasks. On the VQAv2 test-std set, Bridge-Tower achieves an accuracy of $78.73\%$, outperforming the previous state-of-the-art METER model by $1.09\%$ with the same pre-training data and almost no additional parameters and computational cost. Notably, when further scaling the model, Bridge-Tower achieves an accuracy of $81.15\%$, surpassing models that are pre-trained on orders-of-magnitude larger datasets. Code is available at this https URL
Subjects: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL); Machine Learning (cs.LG)
Cite as: arXiv:2206.08657 [cs.CV]
  (or arXiv:2206.08657v1 [cs.CV] for this version)

Submission history

From: Xiao Xu [view email]
[v1] Fri, 17 Jun 2022 09:42:35 GMT (964kb,D)

Link back to: arXiv, form interface, contact.