Kamal Acharya

Ph.D. Candidate at UMBC, AI Researcher

Back to Publications

Conference Paper

SKGHOI: Spatial-Semantic Knowledge Graph for Human-Object Interaction Detection

2023 IEEE ICDMW 2023 DOI: 10.1109/ICDMW60847.2023.00155

Knowledge Graph Human-Object Interaction Computer Vision

DOI Cite

Abstract

This paper presents SKGHOI, a graph-based method for human-object interaction detection that integrates spatial and semantic knowledge. It represents interaction components as graph nodes, captures spatial relationships as edges, combines spatial and semantic encoders, and improves computational efficiency while incorporating prior knowledge for HOI detection.

Plain-Language Summary

The paper improves computer vision models that detect how people interact with objects by using a knowledge graph that captures both spatial relationships and semantic meaning.

Why This Paper Matters

The work offers a more interpretable and efficient graph-based direction for human-object interaction detection.

Research Summary

This paper addresses human-object interaction detection in computer vision. Detecting interactions requires more than recognizing objects; the model must understand relationships between people, objects, spatial context, and semantic meaning.

SKGHOI uses a spatial-semantic knowledge graph to represent interaction components and their relationships. Spatial and semantic encoders are combined to create a richer representation of human-object interactions.

The approach is designed to improve both accuracy and efficiency by incorporating prior knowledge and avoiding overreliance on appearance-only features. This makes it relevant for real-world vision systems that need to understand actions and interactions.

Key Contributions

  • Introduces a spatial-semantic knowledge graph for HOI detection.
  • Combines spatial and semantic encoders to represent interactions.
  • Uses graph structure to incorporate prior knowledge.
  • Demonstrates improved efficiency and performance on HICO-DET.

Publication Details

Type
Conference Paper
Venue
IEEE ICDMW 2023
Year
2023

Authors

Zhu L., Lan Q., Velasquez A., Song H., Kamal A., Tian Q., Niu S.

Research Topics

Knowledge Graph Human-Object Interaction Computer Vision

Citation

@inproceedings{zhu2023-skghoi-spatial-semantic-knowledge-graph-for-human-object-interaction-detection,
  title = {SKGHOI: Spatial-Semantic Knowledge Graph for Human-Object Interaction Detection},
  author = {Zhu L. and Lan Q. and Velasquez A. and Song H. and Kamal A. and Tian Q. and Niu S.},
  booktitle = {IEEE ICDMW 2023},
  year = {2023},
  doi = {10.1109/ICDMW60847.2023.00155}
}