Panoptic Video Scene Graph Generation
This addresses the limitation of existing video scene graph generation in detecting non-rigid objects and backgrounds for comprehensive visual perception systems, though it is incremental as it builds on prior work.
The paper introduces panoptic video scene graph generation (PVSG), a new problem that grounds scene graph nodes with pixel-level segmentation masks instead of bounding boxes to improve video understanding, and contributes a dataset of 400 videos with 150K frames.
Towards building comprehensive real-world visual perception systems, we propose and study a new problem called panoptic scene graph generation (PVSG). PVSG relates to the existing video scene graph generation (VidSGG) problem, which focuses on temporal interactions between humans and objects grounded with bounding boxes in videos. However, the limitation of bounding boxes in detecting non-rigid objects and backgrounds often causes VidSGG to miss key details crucial for comprehensive video understanding. In contrast, PVSG requires nodes in scene graphs to be grounded by more precise, pixel-level segmentation masks, which facilitate holistic scene understanding. To advance research in this new area, we contribute the PVSG dataset, which consists of 400 videos (289 third-person + 111 egocentric videos) with a total of 150K frames labeled with panoptic segmentation masks as well as fine, temporal scene graphs. We also provide a variety of baseline methods and share useful design practices for future work.