English

SEP: Self-Enhanced Prompt Tuning for Visual-Language Model

Computer Vision and Pattern Recognition 2024-11-25 v3

Abstract

Prompt tuning based on Context Optimization (CoOp) effectively adapts visual-language models (VLMs) to downstream tasks by inferring additional learnable prompt tokens. However, these tokens are less discriminative as they are independent of the pre-trained tokens and fail to capture input-specific knowledge, such as class-aware textual or instance-aware visual knowledge. Leveraging the discriminative and generalization capabilities inherent in pre-trained tokens, we introduce a novel approach named Self-Enhanced Prompt Tuning (SEP). The core principle of SEP involves adapting the learnable prompt tokens at each encoder layer from the corresponding self-pretrained tokens, thereby explicitly incorporating discriminative prior knowledge to enhance both textual-level and visual-level embeddings. Furthermore, SEP's self-enhanced tokens not only boost discrimination but also mitigate domain shifts in unseen domains, enhancing generalization. In practice, SEP selects several representative tokens from all pre-trained tokens for each input data at every layer of the text/visual encoders. Subsequently, a Token Fusion Module (TFM) is introduced to generate a self-enhanced token by merging these representative tokens with the learnable tokens using a cross-attention mechanism. This self-enhanced token is then concatenated with all pre-trained tokens, serving as input for subsequent encoder layers to produce the relevant embeddings. Comprehensive evaluations across various benchmarks and tasks confirm SEP's efficacy in prompt tuning. Code: \href{Code}{https://github.com/htyao89/SEP}.

Keywords

Cite

@article{arxiv.2405.15549,
  title  = {SEP: Self-Enhanced Prompt Tuning for Visual-Language Model},
  author = {Hantao Yao and Rui Zhang and Lu Yu and Yongdong Zhang and Changsheng Xu},
  journal= {arXiv preprint arXiv:2405.15549},
  year   = {2024}
}
R2 v1 2026-06-28T16:38:56.451Z