BridgeTower: Building Bridges Between Encoders in Vision-Language Representation Learning
Xu, Xiao, Wu, Chenfei, Rosenman, Shachar, Lal, Vasudev, Che, Wanxiang, Duan, Nan
–arXiv.org Artificial Intelligence
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 deep cross-modal encoder, or feed the last-layer uni-modal representations from the deep pre-trained uni-modal encoders into the top cross-modal encoder. Both approaches potentially restrict vision-language representation learning and limit model performance. In this paper, we propose BridgeTower, which introduces 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 effective bottom-up cross-modal alignment and fusion between visual and textual representations of different semantic levels of pre-trained uni-modal encoders in the cross-modal encoder. Pre-trained with only 4M images, BridgeTower achieves state-of-the-art performance on various downstream vision-language tasks. In particular, on the VQAv2 test-std set, BridgeTower achieves an accuracy of 78.73%, outperforming the previous state-of-the-art model METER by 1.09% with the same pre-training data and almost negligible additional parameters and computational costs. Notably, when further scaling the model, BridgeTower achieves an accuracy of 81.15%, surpassing models that are pre-trained on orders-of-magnitude larger datasets. Code and checkpoints are available at https://github.com/microsoft/BridgeTower.
arXiv.org Artificial Intelligence
Jun-9-2023
- Country:
- South America > Chile
- Oceania > Australia
- North America
- Mexico (0.04)
- United States
- Minnesota > Hennepin County
- Minneapolis (0.14)
- Nevada > Clark County
- Las Vegas (0.04)
- Massachusetts > Suffolk County
- Boston (0.04)
- Hawaii > Honolulu County
- Honolulu (0.04)
- Louisiana > Orleans Parish
- New Orleans (0.04)
- Utah > Salt Lake County
- Salt Lake City (0.04)
- Washington > King County
- Seattle (0.04)
- California
- Los Angeles County > Long Beach (0.04)
- San Diego County > San Diego (0.04)
- New York > New York County
- New York City (0.04)
- Minnesota > Hennepin County
- Canada
- Quebec > Montreal (0.04)
- British Columbia > Metro Vancouver Regional District
- Vancouver (0.14)
- Europe
- Germany > Berlin (0.04)
- Austria (0.04)
- Switzerland > Zürich
- Zürich (0.14)
- Netherlands > North Holland
- Amsterdam (0.04)
- Italy > Tuscany
- Florence (0.04)
- Denmark > Capital Region
- Copenhagen (0.04)
- Spain > Andalusia
- Granada Province > Granada (0.04)
- Ukraine > Kyiv Oblast
- Kyiv (0.04)
- Ireland > Leinster
- County Dublin > Dublin (0.04)
- Belgium > Brussels-Capital Region
- Brussels (0.04)
- Asia > China
- Yunnan Province > Kunming (0.04)
- Shandong Province > Qingdao (0.04)
- Hong Kong (0.04)
- Heilongjiang Province > Harbin (0.04)
- Africa > Ethiopia
- Addis Ababa > Addis Ababa (0.04)
- Genre:
- Research Report (0.84)
- Technology: