Computer Vision to Secure Your Surroundings with AI/ML Solution Built Using Open Source Tools at the Edge

Developing ethical AI software for edge devices

• AI software can be used for good or bad, so it's important to have a strong moral compass and principles guiding development
• Intel abides by six ethics principles, including respecting human rights and enabling human oversight
• Challenges in deploying AI at the edge include resource constraints and data management
• Postgres database was chosen for data management due to its reliability, efficiency, and security features
• Security measures were taken to ensure sensitive information is not accessible to unauthorized users

The speaker mentioned that AI software can be used for good or bad, and it's important to have a strong moral compass when developing it. They also shared that Intel abides by six ethics principles to ensure their AI software is used for good. The speaker discussed challenges in deploying AI at the edge, including resource constraints and data management. They explained that they chose to use a Postgres database for data management due to its reliability, efficiency, and security features. Finally, the speaker emphasized the importance of security measures to ensure sensitive information is not accessible to unauthorized users.

Fast, low-cost edge compute is supporting the growth of IoT, AI and Computer Vision (CV) based solutions in several fields including smart city/home. Security solutions aiding situational awareness have benefits in keeping assets and the public safe. However, these solutions are increasingly difficult to develop & deploy due to resource constraints, hardware costs, and high inference loads on the edge device. Our team developed a CV based Security as a Service Smart City Solution using AI/ML. This solution provides a framework and processing pipeline for deploying AI-assisted, multi-camera Smart City Solution of vehicular and walkway traffic. The Open Source software leveraged includes: GStreamer multimedia framework, Intel Distribution of OpenVINO Toolkit, Angular UI using VideoJS, Grafana map to depict edge device locations, PostgreSQL, and EdgeX Foundry as an optional listener for inference results. The solution uses a containerized microservice-based architecture. This presentation walks through the learnings and challenges encountered during the design and implementation of this unique solution for AI at the edge for a Security as a Service solution. We will also discuss the ethical concerns that drive our moral compass in developing these types of CV solutions.