Building a CCTV Vision pipeline for AI based home automation using Frigate

Aceptada

This talk addresses the complexities of building a CCTV vision pipeline for AI-based home automation. The speaker will share their process of using Frigate, Nvidia Jetson, and custom code to create this pipeline. The resulting system enables novel applications like a virtual dog and parking assistant, demonstrating practical AI vision in home automation.

Tipo: Talk / Charla

Nivel: Medio

Disertantes: Matias Cuenca-Acuna

Biografiá del Disertante: Matias Cuenca-Acuna, a Principal Engineer/Software Architect, has been providing infrastructural software solutions for over 30 years. His skills include distributed systems, security operations, Internet of Things, and data analysis. At LandingAI, he focuses on solving computer vision problems by ensuring their visual AI solutions are cloud-compatible, scalable, highly available, compliant, and secure. Matias's background includes a PhD in Computer Science from Rutgers University and a history of published work in distributed computing and security.

Horario: 12:00 - 13:00 - 26/04/2025

Sala: Aula 3

Etiquetas/Tags: machine learning visual ai frigate

Descripción completa

While obtaining a "decent" machine learning model has become significantly easier, applying AI to real-world vision for home automation presents substantial hurdles that extend far beyond model development. This talk will explore the complexities of building a robust CCTV vision pipeline for AI-powered home automation. We will delve into the critical challenges involved in creating such a system, including temporal continuity, compute resources management, resiliency, etc. In this session, I will share my experience in overcoming these challenges by constructing a functional vision pipeline using Frigate, an open-source NVR, alongside NVidia's Jetson platform and custom-developed code. The talk will highlight how these technologies can be combined to enable novel home automation use cases, such as a virtual dog and a parking assistant, illustrating the practical application of AI vision in a home environment.