OpenPSG: Open-set Panoptic Scene Graph Generation via Large Multimodal Models

Zijian Zhou,Zheng Zhu,Holger Caesar,Miaojing Shi

2024-07-16

Abstract:Panoptic Scene Graph Generation (PSG) aims to segment objects and recognize their relations, enabling the structured understanding of an image. Previous methods focus on predicting predefined object and relation categories, hence limiting their applications in the open world scenarios. With the rapid development of large multimodal models (LMMs), significant progress has been made in open-set object detection and segmentation, yet open-set relation prediction in PSG remains unexplored. In this paper, we focus on the task of open-set relation prediction integrated with a pretrained open-set panoptic segmentation model to achieve true open-set panoptic scene graph generation (OpenPSG). Our OpenPSG leverages LMMs to achieve open-set relation prediction in an autoregressive manner. We introduce a relation query transformer to efficiently extract visual features of object pairs and estimate the existence of relations between them. The latter can enhance the prediction efficiency by filtering irrelevant pairs. Finally, we design the generation and judgement instructions to perform open-set relation prediction in PSG autoregressively. To our knowledge, we are the first to propose the open-set PSG task. Extensive experiments demonstrate that our method achieves state-of-the-art performance in open-set relation prediction and panoptic scene graph generation. Code is available at \url{<a class="link-external link-https" href="https://github.com/franciszzj/OpenPSG" rel="external noopener nofollow">this https URL</a>}.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

### Problems Addressed by the Paper

This paper aims to address the issue of open-set relationship prediction in Panoptic Scene Graph Generation (PSG). Traditional PSG methods can only predict predefined object and relationship categories, which limits their application in open-world scenarios. With the development of Large Multimodal Models (LMMs), significant progress has been made in open-set object detection and segmentation, but open-set relationship prediction in PSG has not been fully explored.

The main contributions of the paper are as follows:
1. **Proposing the Open-Set PSG Task**: Combining pre-trained open-set panoptic segmentation models to achieve true open-set panoptic scene graph generation (OpenPSG).
2. **Utilizing LMMs for Open-Set Relationship Prediction**: Employing LMMs in an autoregressive manner to achieve open-set relationship prediction.
3. **Introducing the Relationship Query Transformer**: Efficiently extracting visual features of object pairs and estimating the existence of relationships between them.
4. **Designing Generation and Judgment Instructions**: Executing open-set relationship prediction in PSG.

Experimental results show that this method achieves state-of-the-art performance in open-set relationship prediction and panoptic scene graph generation.