Conditional Prompt Learning for Vision-Language Models

TL;DR

This paper introduces CoCoOp, a dynamic prompt learning method for vision-language models that improves generalization to unseen classes and domains by generating input-conditional prompts, addressing overfitting issues of previous static prompt methods.

Contribution

The paper proposes Conditional Context Optimization (CoCoOp), a novel dynamic prompt learning approach that enhances generalization and transferability of vision-language models to unseen classes and domains.

Findings

CoCoOp outperforms CoOp on unseen classes within the same dataset.

CoCoOp demonstrates better transferability across different datasets.

The method improves domain generalization performance.

Abstract

With the rise of powerful pre-trained vision-language models like CLIP, it becomes essential to investigate ways to adapt these models to downstream datasets. A recently proposed method named Context Optimization (CoOp) introduces the concept of prompt learning -- a recent trend in NLP -- to the vision domain for adapting pre-trained vision-language models. Specifically, CoOp turns context words in a prompt into a set of learnable vectors and, with only a few labeled images for learning, can achieve huge improvements over intensively-tuned manual prompts. In our study we identify a critical problem of CoOp: the learned context is not generalizable to wider unseen classes within the same dataset, suggesting that CoOp overfits base classes observed during training. To address the problem, we propose Conditional Context Optimization (CoCoOp), which extends CoOp by further learning a lightweight neural network to generate for each image an input-conditional token (vector). Compared to CoOp's static prompts, our dynamic prompts adapt to each instance and are thus less sensitive to class shift. Extensive experiments show that CoCoOp generalizes much better than CoOp to unseen classes, even showing promising transferability beyond a single dataset; and yields stronger domain generalization performance as well. Code is available at https://github.com/KaiyangZhou/CoOp.