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

TL;DR

FIBER introduces a novel vision-language model that integrates deep multimodal fusion within the backbone and employs a two-stage pre-training strategy, achieving superior performance across diverse VL tasks with efficient data usage.

Contribution

The paper proposes FIBER, a unified VL model with fusion embedded in the backbone and a two-stage pre-training approach, enabling it to excel in both high-level and region-level understanding tasks.

Findings

FIBER outperforms strong baselines on multiple VL tasks.

Deep fusion in the backbone improves memory efficiency and performance.

Two-stage pre-training effectively leverages coarse and fine-grained data.

Abstract

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, bringing gains in terms of memory and performance. 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 available at https://github.com/microsoft/FIBER.