ST-Adapter: Parameter-Efficient Image-to-Video Transfer Learning

TL;DR

This paper introduces ST-Adapter, a parameter-efficient method for adapting large pre-trained image models to video understanding tasks, achieving competitive performance with significantly fewer trainable parameters.

Contribution

The work proposes a novel spatio-temporal adapter for cross-modality transfer learning from images to videos, enabling efficient fine-tuning with minimal parameter updates.

Findings

Matches or outperforms full fine-tuning and state-of-the-art video models.

Requires approximately 8% of parameters per task compared to previous methods.

Demonstrates effective reasoning about dynamic video content using a compact design.

Abstract

Capitalizing on large pre-trained models for various downstream tasks of interest have recently emerged with promising performance. Due to the ever-growing model size, the standard full fine-tuning based task adaptation strategy becomes prohibitively costly in terms of model training and storage. This has led to a new research direction in parameter-efficient transfer learning. However, existing attempts typically focus on downstream tasks from the same modality (e.g., image understanding) of the pre-trained model. This creates a limit because in some specific modalities, (e.g., video understanding) such a strong pre-trained model with sufficient knowledge is less or not available. In this work, we investigate such a novel cross-modality transfer learning setting, namely parameter-efficient image-to-video transfer learning. To solve this problem, we propose a new Spatio-Temporal Adapter (ST-Adapter) for parameter-efficient fine-tuning per video task. With a built-in spatio-temporal reasoning capability in a compact design, ST-Adapter enables a pre-trained image model without temporal knowledge to reason about dynamic video content at a small (~8%) per-task parameter cost, requiring approximately 20 times fewer updated parameters compared to previous work. Extensive experiments on video action recognition tasks show that our ST-Adapter can match or even outperform the strong full fine-tuning strategy and state-of-the-art video models, whilst enjoying the advantage of parameter efficiency. The code and model are available at https://github.com/linziyi96/st-adapter