computer-vision-5

Panoptic Scene Graph Generation

3260 papers • 126 benchmarks • 313 datasets

PSG task abstracts the given image with a scene graph, where nodes are grounded by panoptic segmentation

(Image credit: Papersgraph)

Benchmarks

These leaderboards are used to track progress in panoptic-scene-graph-generation-5

Trend

Dataset

Best Model

Actions

PSG Dataset

Libraries

i

Use these libraries to find panoptic-scene-graph-generation-5 models and implementations

shikorab/SceneGraph

2 papers 69

Datasets

PSG Dataset

Subtasks

No subtasks available.

Most implemented papers

Neural Motifs: Scene Graph Parsing with Global Context

Rowan Zellers, Yejin Choi, Mark Yatskar, Sam Thomson•Thu Nov 16 2017

This work analyzes the role of motifs: regularly appearing substructures in scene graphs and introduces Stacked Motif Networks, a new architecture designed to capture higher order motifs in scene graph graphs that improves on the previous state-of-the-art by an average of 3.6% relative improvement across evaluation settings.

1105

Content

0

Paper Graph

Learning to Compose Dynamic Tree Structures for Visual Contexts

Wenhan Luo, Baoyuan Wu, W. Liu, Kaihua Tang, Hanwang Zhang•Tue Dec 04 2018

A hybrid learning procedure is developed which integrates end-task supervised learning and the tree structure reinforcement learning, where the former's evaluation result serves as a self-critic for the latter's structure exploration.

562 0

Paper Graph

Scene Graph Generation by Iterative Message Passing

Li Fei-Fei, Yuke Zhu, C. Choy, Danfei Xu•Mon Jan 09 2017

This work explicitly model the objects and their relationships using scene graphs, a visually-grounded graphical structure of an image, and proposes a novel end-to-end model that generates such structured scene representation from an input image.

1336 0

Paper Graph

Panoptic Video Scene Graph Generation

Ziwei Liu, Kaiyang Zhou, Chen Change Loy, Wayne Zhang, Bo Li, Liangyu Chen, Jingkang Yang, Zujin Guo, Wen-Hsiao Peng, Xiangtai Li, Zheng Ma•Wed May 31 2023

A high-quality PVSG dataset is contributed, which consists of 400 videos (289 third-person + 111 egocentric videos) with totally 150K frames labeled with panoptic segmentation masks as well as fine, temporal scene graphs.

55 0

Paper Graph

HiLo: Exploiting High Low Frequency Relations for Unbiased Panoptic Scene Graph Generation

Holger Caesar, Miaojing Shi, Zijian Zhou•Mon Mar 27 2023

The proposed HiLo framework lets different network branches specialize on low and high frequency relations, enforce their consistency and fuse the results, and is the first to propose an explicitly unbiased Panoptic Scene Graph generation method.

28 0

Paper Graph

Panoptic Scene Graph Generation

Kaiyang Zhou, Wayne Zhang, Ziwei Liu, Jingkang Yang, Yi Zhe Ang, Zujin Guo•Thu Jul 21 2022

Panoptic scene graph generation (PSG) is introduced, a new problem task that requires the model to generate a more comprehensive scene graph representation based on panoptic segmentations rather than rigid bounding boxes.

148 0

Paper Graph

Pair Then Relation: Pair-Net for Panoptic Scene Graph Generation

Ziwei Liu, Jinghao Wang, Jingkang Yang, Zujin Guo, Xiangtai Li, Zhengyu Wen•Sun Jul 16 2023

A novel framework is presented: Pair then Relation (Pair-Net), which uses a Pair Proposal Network (PPN) to learn and filter sparse pair-wise relationships between subjects and objects and achieves over 10% absolute gains compared to the baseline, PSGFormer.

28 0

Paper Graph

VLPrompt: Vision-Language Prompting for Panoptic Scene Graph Generation

Zijian Zhou, Miaojing Shi, Holger Caesar•Sun Nov 26 2023

The Vision-Language Prompting (VLPrompt) model is proposed, which acquires vision information from images and language information from LLMs, and through a prompter network based on attention mechanism, it achieves precise relation prediction.

15 0

Paper Graph

Panoptic Scene Graph Generation with Semantics-prototype Learning

Li Li, Roger Zimmermann, Wei Ji, Yiming Wu, Meng Li, Youxuan Qin, Lina Wei•Thu Jul 27 2023

A novel framework named ADTrans is proposed to adaptively transfer biased predicate annotations to informative and unified ones, and to promise consistency and accuracy during the transfer process, and learn unbiased prototypes of predicates with different intensities.

54 0

Paper Graph

Adding a benchmark result helps the community track progress.