Experts from the analytical company ABI Research believe that the base of smart cameras, endowed with artificial intelligence (AI), installed around the world, in 2025 will exceed 350 million pieces. The use of AI chipsets in CCTV cameras for smart cities will by then become the norm.
More than 65% of cameras shipped in 2025 are expected to have at least one AI chip. These cameras will be equipped with deep learning (DL) models to automate and expand decision-making in applications such as intelligent traffic management, pedestrian surveillance, pedestrian traffic management, and security, including early detection of threats.
Currently, most of these workloads are performed either by cloud-hosted GO models offered by video analytics vendors such as SenseTime, Ipsotek, Vicentina, and Sentry AI, or by smart cameras and video recorders from companies such as HikVision and Dahua. Analysts acknowledge that both deployment methods have strengths and weaknesses.
“Join our channel on Telegram”
Additionally, there will be two technology trends that will catalyze the deployment of deep learning-based machine vision.
The first is edge computing, which allows, instead of deploying certain HE models in smart cameras, which are several times more expensive than conventional cameras, to place HE models on gateways and local servers. This allows data to be processed and stored at the edge with faster response times compared to cloud infrastructure.
The second is the adoption of 5G. Although the network segment will not be commercially ready by 2023, the ability to segment the 5G network allows service providers to offer dedicated network resources to host microservices, provide high service reliability, and seamless device connectivity – all well suited for GO-based machine vision in a smart city.
Of course, there are also obstacles. Analysts include distrust of a part of society and the rules associated with the introduction of AI into cameras to them. Human rights defenders around the world fear abuse and oppose the introduction of facial recognition technologies. ABI Research believes that technology vendors can be successful if they demonstrate transparent and understandable civil society models and a willingness to embrace open and shared standards and ethics.
