Explainable Deep Networks

Abstract

As recently as December of 2016, terrorist groups such as ISIS have called for an increase in attacks against US Naval Bases. To prevent such incidents, video surveillance cameras can be used to monitor critical areas for potential threats. However, monitoring video surveillance cameras can be a challenging task. Atypical setup employs many cameras which all produce a significant amount of information. The manpower required to monitor these cameras is tremendous, which can then lead to a reliance on automated video analytics. At issue is the effectiveness of these analytics in communicating important information to the warfighter. State-of-the-art approaches can locate potential threats, but are unable to explain why something might be a threat. The warfighter needs actionable intelligence: explanations are important as they can provide this important information.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Aug 30, 2021
Accession Number
AD1149306

Entities

People

  • Wallace Lawson

Organizations

  • United States Naval Research Laboratory

Tags

Communities of Interest

  • Autonomy
  • Human Systems

DTIC Thesaurus Topics

  • Anomaly Detection
  • Artificial Intelligence
  • Artificial Intelligence Software
  • Case Studies
  • Computer Vision
  • Computers
  • Convolutional Neural Networks
  • Deep Learning
  • Detection
  • Information Processing
  • Information Systems
  • Instructors
  • Neural Networks
  • Object Recognition
  • Ontologies
  • Recognition
  • Surveillance
  • Video Surveillance

Fields of Study

  • Computer science

Readers

  • Computer Vision.
  • Distributed Systems and Data Platform Development
  • Political Violence and Terrorism Studies.