Set-of-Mark Prompting Unleashes Extraordinary Visual Grounding in GPT-4V

TL;DR

Set-of-Mark prompting significantly enhances GPT-4V's ability to perform fine-grained visual grounding by using image segmentation and marking techniques, leading to state-of-the-art results in various multimodal tasks.

Contribution

Introduces Set-of-Mark (SoM), a novel visual prompting method that leverages off-the-shelf segmentation models to improve multimodal reasoning in GPT-4V.

Findings

Outperforms state-of-the-art in referring expression comprehension

Effective in zero-shot visual grounding tasks

Validates across diverse vision and multimodal benchmarks

Abstract

We present Set-of-Mark (SoM), a new visual prompting method, to unleash the visual grounding abilities of large multimodal models (LMMs), such as GPT-4V. As illustrated in Fig. 1 (right), we employ off-the-shelf interactive segmentation models, such as SEEM/SAM, to partition an image into regions at different levels of granularity, and overlay these regions with a set of marks e.g., alphanumerics, masks, boxes. Using the marked image as input, GPT-4V can answer the questions that require visual grounding. We perform a comprehensive empirical study to validate the effectiveness of SoM on a wide range of fine-grained vision and multimodal tasks. For example, our experiments show that GPT-4V with SoM in zero-shot setting outperforms the state-of-the-art fully-finetuned referring expression comprehension and segmentation model on RefCOCOg. Code for SoM prompting is made public at: https://github.com/microsoft/SoM.