Set-of-Mark Prompting Unleashes Extraordinary Visual Grounding in GPT-4V
Yang, Jianwei, Zhang, Hao, Li, Feng, Zou, Xueyan, Li, Chunyuan, Gao, Jianfeng
–arXiv.org Artificial Intelligence
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.
arXiv.org Artificial Intelligence
Nov-6-2023
- Country:
- North America > United States > Wisconsin (0.14)
- Genre:
- Research Report (1.00)
- Industry:
- Health & Medicine > Consumer Health (1.00)
- Leisure & Entertainment > Games
- Computer Games (0.68)
- Technology: