Dynamic Scene Graphs for Extracting Activity-based Intelligence

Abstract

Artificial intelligence and machine learning research have made significant contributions that exploit machine intelligence to support Activity-based Intelligence (ABI) extraction in many military operations. However,ABI extraction in open-world scenarios poses a variety of technical difficulties that have not been fully addressed.For example, as the battlefield is a complex and highly dynamic environment, novel objects and events will frequently appear and it is necessary to understand these novelties. In addition, while good accuracy can often be achieved using state-of-the-art deep learning-based approaches, their poor explainability makes the extracted information hard to be used by the human-decision team. Furthermore, it is also challenging to incorporate human knowledge to further enhance the performance of machine learning approaches.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Apr 26, 2023
Accession Number
AD1226854

Entities

People

  • Qi Yu
  • Yu Kong

Organizations

  • Rochester Institute of Technology

Tags

Fields of Study

  • Computer science

Readers

  • Geospatial Intelligence and Artificial Intelligence Analytics
  • Neural Network Machine Learning.
  • Team-Based Human-Centered Cognitive Task Decision Making and Information Performance.

Technology Areas

  • AI & ML
  • AI & ML - DoD AI Strategy
  • AI & ML - Neural Networks