Skip to yearly menu bar Skip to main content


Coarse-to-Fine Vision-Language Pre-training with Fusion in the Backbone

Zi-Yi Dou · Aishwarya Kamath · Zhe Gan · Pengchuan Zhang · Jianfeng Wang · Linjie Li · Zicheng Liu · Ce Liu · Yann LeCun · Nanyun Peng · Jianfeng Gao · Lijuan Wang

Hall J (level 1) #635

Keywords: [ VQA ] [ Object Detection ] [ vision-language pre-training ] [ image captioning ]


Vision-language (VL) pre-training has recently received considerable attention. However, most existing end-to-end pre-training approaches either only aim to tackle VL tasks such as image-text retrieval, visual question answering (VQA) and image captioning that test high-level understanding of images, or only target region-level understanding for tasks such as phrase grounding and object detection. We present FIBER (Fusion-In-the-Backbone-based transformER), a new VL model architecture that can seamlessly handle both these types of tasks. Instead of having dedicated transformer layers for fusion after the uni-modal backbones, FIBER pushes multimodal fusion deep into the model by inserting cross-attention into the image and text backbones to better capture multimodal interactions. In addition, unlike previous work that is either only pre-trained on image-text data or on fine-grained data with box-level annotations, we present a two-stage pre-training strategy that uses both these kinds of data efficiently: (i) coarse-grained pre-training based on image-text data; followed by (ii) fine-grained pre-training based on image-text-box data. We conduct comprehensive experiments on a wide range of VL tasks, ranging from VQA, image captioning, and retrieval, to phrase grounding, referring expression comprehension, and object detection. Using deep multimodal fusion coupled with the two-stage pre-training, FIBER provides consistent performance improvements over strong baselines across all tasks, often outperforming methods using magnitudes more data. Code is released at

Chat is not available.