Vision Transformer Adapter for Dense Predictions

TL;DR

This paper introduces ViT-Adapter, a simple module that enhances plain Vision Transformers for dense prediction tasks, achieving state-of-the-art results without additional training data or vision-specific inductive biases.

Contribution

The paper proposes ViT-Adapter, enabling plain ViT models to perform competitively on dense prediction tasks by incorporating a pre-training-free, image-related inductive bias.

Findings

Achieved 60.9 box AP and 53.0 mask AP on COCO test-dev.

ViT-Adapter matches performance of vision-specific transformers.

No extra detection data needed for state-of-the-art results.

Abstract

This work investigates a simple yet powerful dense prediction task adapter for Vision Transformer (ViT). Unlike recently advanced variants that incorporate vision-specific inductive biases into their architectures, the plain ViT suffers inferior performance on dense predictions due to weak prior assumptions. To address this issue, we propose the ViT-Adapter, which allows plain ViT to achieve comparable performance to vision-specific transformers. Specifically, the backbone in our framework is a plain ViT that can learn powerful representations from large-scale multi-modal data. When transferring to downstream tasks, a pre-training-free adapter is used to introduce the image-related inductive biases into the model, making it suitable for these tasks. We verify ViT-Adapter on multiple dense prediction tasks, including object detection, instance segmentation, and semantic segmentation. Notably, without using extra detection data, our ViT-Adapter-L yields state-of-the-art 60.9 box AP and 53.0 mask AP on COCO test-dev. We hope that the ViT-Adapter could serve as an alternative for vision-specific transformers and facilitate future research. The code and models will be released at https://github.com/czczup/ViT-Adapter.