Research map / Video / Video segmentation and tracking
Panoptic Video Scene Graph Generation
Jingkang Yang, Wenxuan Peng, Xiangtai Li, Zujin Guo, Liangyu Chen, Bo Li, Zheng Ma, Kaiyang Zhou, Wayne Zhang, Chen Change Loy, Ziwei Liu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use a Transformer Encoder instead of Vanilla relation prediction (ablation) | IPS+T · PVSG test, thre = 0.5 | mR@20: 1.35 -> 2.81 | Improved |
| Use 1D Convolution instead of Vanilla relation prediction (ablation) | IPS+T · PVSG test, thre = 0.5 | R@20: 3.04 -> 3.88 | Improved |
| Replace the Handcrafted Window with 1D Convolution (ablation) | IPS+T · PVSG test, thre = 0.5 | R@20: 2.52 -> 3.88 | Improved |
| Use Transformer Encoder instead of 1D Convolution (ablation) | IPS+T · PVSG test, thre = 0.5 | mR@20: 2.55 -> 2.81 | Improved |
| Use Transformer Encoder instead of Vanilla relation prediction (ablation) | VPS · PVSG test, thre = 0.5 | mR@20: 0.10 -> 0.61 | Improved |
| Use 1D Convolution instead of Vanilla relation prediction (ablation) | VPS · PVSG test, thre = 0.1 | R@20: 6.29 -> 8.07 | Improved |
| Use 1D Convolution instead of the Handcrafted Window (ablation) | VPS · PVSG test, thre = 0.1 | mR@20: 2.84 -> 7.84 | Improved |
| Use Transformer Encoder instead of 1D Convolution (ablation) | VPS · PVSG test, thre = 0.1 | R@20: 8.07 -> 6.50 | Hurt |
| Use the IPS+T first stage rather than the end-to-end VPS first stage | Transformer Encoder relation model · PVSG test, thre = 0.5 | R@20: 0.42 -> 3.88 | Improved |
| Use the IPS+T first stage rather than the end-to-end VPS first stage | 1D Convolution relation model · PVSG test, thre = 0.1 | R@20: 8.07 -> 10.06 | Improved |
| Relax the volume-overlap threshold from 0.5 to 0.1 | IPS+T + 1D Convolution · PVSG test | R@20: 3.88 -> 10.06 | Improved |
Closest papers by meaning
- Towards Unbiased and Robust Spatio-Temporal Scene Graph Generation and Anticipation (CVPR 2025 · similarity 0.87)
- DiffVsgg: Diffusion-Driven Online Video Scene Graph Generation (CVPR 2025 · similarity 0.87)
- A Closer Look at Dynamic Scene Graph Generation in the Era of Multimodal Large Language Models (NeurIPS 2026 · similarity 0.87)
- Motion-aware Contrastive Learning for Temporal Panoptic Scene Graph Generation (AAAI 2025 · similarity 0.87)
- OED: Towards One-stage End-to-End Dynamic Scene Graph Generation (CVPR 2024 · similarity 0.87)
- DSGG: Dense Relation Transformer for an End-to-end Scene Graph Generation (CVPR 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
