TuneVLSeg: Prompt Tuning Benchmark for Vision-Language Segmentation Models

TL;DR

This paper introduces TuneVLSeg, a benchmarking framework for prompt tuning in vision-language segmentation models, evaluating various prompt strategies across diverse datasets, and highlighting the effectiveness of visual prompts over textual ones under domain shifts.

Contribution

It presents a comprehensive benchmark for prompt tuning in VLSMs, including multiple strategies and datasets, and provides insights into their performance under domain shifts.

Findings

Visual prompt tuning often outperforms textual prompt tuning.

Multimodal prompt tuning has comparable performance with fewer hyperparameters.

Prompt tuning effectiveness varies significantly across domain shifts.

Abstract

Vision-Language Models (VLMs) have shown impressive performance in vision tasks, but adapting them to new domains often requires expensive fine-tuning. Prompt tuning techniques, including textual, visual, and multimodal prompting, offer efficient alternatives by leveraging learnable prompts. However, their application to Vision-Language Segmentation Models (VLSMs) and evaluation under significant domain shifts remain unexplored. This work presents an open-source benchmarking framework, TuneVLSeg, to integrate various unimodal and multimodal prompt tuning techniques into VLSMs, making prompt tuning usable for downstream segmentation datasets with any number of classes. TuneVLSeg includes $6$ prompt tuning strategies on various prompt depths used in $2$ VLSMs totaling of $8$ different combinations. We test various prompt tuning on $8$ diverse medical datasets, including $3$ radiology datasets (breast tumor, echocardiograph, chest X-ray pathologies) and $5$ non-radiology datasets (polyp, ulcer, skin cancer), and two natural domain segmentation datasets. Our study found that textual prompt tuning struggles under significant domain shifts, from natural-domain images to medical data. Furthermore, visual prompt tuning, with fewer hyperparameters than multimodal prompt tuning, often achieves performance competitive to multimodal approaches, making it a valuable first attempt. Our work advances the understanding and applicability of different prompt-tuning techniques for robust domain-specific segmentation. The source code is available at https://github.com/naamiinepal/tunevlseg.