Out of the 117 million cameras shipped worldwide in 2020, only 16% had deep learning-based analytics functionality. The remaining 84% of sold cameras present several limitations and risks for their owners.

Data Security

Without onboard analytics, cameras must transmit their video streams to a server for subsequent analysis and storage. This increases the number of points and the volume of data vulnerable to security breaches. Each additional transmission point adds another potential weakness where data can be intercepted or compromised.

Increased Network Load

Server-based analysis requires the transmission of all camera data to a central data processing center. This creates a much greater demand for an expensive, high-bandwidth network infrastructure. The need to handle large volumes of data can strain network resources, leading to potential bottlenecks and increased operational costs.

Higher Costs for Power and Data Processing

In some scenarios, the need to transmit data across the network for centralized processing can be entirely avoided, but in others, the sheer volume of data involved makes centralized processing impractical. Even with advanced compression codecs, the increased number of cameras and higher resolutions exacerbate the issue. Streaming large amounts of video and then analyzing it using server-based GPUs requires significant financial investment. By contrast, when data analysis occurs directly on the camera, the constant streaming of video and the need for an extensive infrastructure are minimized.

Data Transmission and Processing Delays

Real-time scenarios, such as facial recognition for access control or sending immediate alerts, particularly in critical situations, benefit greatly fr om image analysis directly on the camera. When cameras handle processing onboard, they eliminate the delays associated with data transmission to a server and back. This is crucial for applications wh ere immediate response times are essential.

For tasks like tracking objects or individuals over large areas using multiple cameras, the advantages of onboard deep learning-based analytics become clear. These systems benefit from the enhanced computational power available on the cameras, reducing latency and the time required to process video data.

Additionally, some environments may not have critical processes but still require a smooth user experience. For example, in retail POS systems or self-checkout kiosks, any delay in system response can render them practically unusable, especially if internet speeds drop or connectivity is lost. Quick and efficient processing on the camera itself ensures that such systems remain responsive and functional, providing a seamless experience for users.