Required Hardware
This page contains information about hardware, general hardware requirements and sample hardware for particular usecases.
General Hardware requirements
To install SmartFace Platform successfully, your Windows or Linux hardware must meet the following general requirements:
- x86_64 or ARM CPU
- CPU with support for AVX2 instruction set
- CPU with at least 4 physical cores
- 16GB RAM
- 80 GB of storage
Evaluation of SmartFace
The best evaluation experience of SmartFace can only be achieved with a propper hardware. Underestimating the hardware needs results in a poor user experience, including latency, timeouts and delays. Linux is preferred operating system. For the seamless installation process follow our Linux Installation guide.
The table below provides the hardware employed by the SmartFace team for the purposes of demonstration and showcasing.
Hardware | RTSP Cameras | Video Files |
---|---|---|
NUC with Intel Core i7 1360P Raptor Lake 5 GHz, 16 GB RAM | up to 3 cameras | 2x speed up for 1 video file |
Lenovo Legion 9 with Intel Core i9 14900HX, 64GB RAM | up to 18 cameras | 10x speed up for 1 video file |
DELL PowerEdge R250 with Xeon E 2336, 32 GB RAM | up to 4 cameras | 2x speed up for 1 video file |
Production Deployment
This topic describes sample hardware setup for four common scenarios of using SmartFace. Each of provided hardware sample includes estimated processing capabilities.
Conditions to meet to achieve estimated performance are:
- RTSP resolution is set to Full HD
- Number of registered watchlist members is less than 10k
Video Surveillance
The Video Surveillance setup is tailored for processing RTSP video streams from IP cameras. Processing is performed on the server side, therefore the server hardware requirements are addequate.
In the table below, the processing capabilities are listed for several server configurations, including RTSP in Full HD and a Watchlist size of 10k.
Hardware | RTSP Cameras |
---|---|
DELL PowerEdge R250 with Intel Xeon E 2336, 32 GB RAM | up to 4 cameras |
DELL PowerEdge R660 with Intel Xeon Gold 5416S | up to 14 cameras |
DELL PowerEdge R760 with 2x Intel Xeon Platinum 8470 | up to 100 cameras |
Lightweight Facial Identification Service
The Lightweight Facial Identification Service setup is all about the API and request/response traffic, running either on premise or in public available cloud (fully supported is Amazon AWS, Microsoft Azure or Google Cloud Platform).
The table below describes Search request capabilities with throughput per second and with response latency under 1000ms.
Hardware | Search requests | Watchlist Size |
---|---|---|
NUC with Intel Core i7 1360P Raptor Lake 5 GHz, 16 GB RAM | 4 requests per second | up to 100k |
Amazon EC2 c7g.2xlarge | 2 requests per second | up to 1M |
Amazon EC2 c7g.4xlarge | 14 requests per second | up to 1M |
2x Amazon EC2 c7g.12xlarge | 120 requests per second | up to 1M |
- search request contains selfie-like image with one person
Edge Processing
The Edge Processing setup is tailored for Smart Cameras and Edge Devices, where all the heavy lifting is performed on the edge (or near the edge). The same hardware used for conventional RTSP processing can efficiently handle significantly higher traffic volumes.
Hardware | Edge Devices | Watchlist Size |
---|---|---|
NUC with Intel Core i7 1360P Raptor Lake 5 GHz, 16 GB RAM | up to 150 edge devices | up to 10k people |
DELL PowerEdge R250 with Xeon E 2336, 32 GB RAM | up to 150 edge devices | up to 10k people |
DELL PowerEdge R760 with 2x Intel Xeon Platinum 8470 | up to 400 edge devices | up to 10M people |
Supported Smart Cameras and Edge devices
SmartFace Embedded currently supports IP cameras using Ambarella SoC, AI boxes using Hailo chip or NVidia Jetson based devices.
Refer to list of Supported devices for more information.
Video Investigation
In the Video Investigation scenario, all decoding, detection, and analytics tasks are performed on the server side. Depending on the desired speed expectations, the requirements for server hardware need to be appropriately scaled.
The table below lists the processing capabilities of several server configurations for the processing of 10 hours of video.
Hardware | Video Files |
---|---|
NUC with Intel Core i7 1360P Raptor Lake 5 GHz, 16 GB RAM | 2x speed up - processed in 5 hours |
DELL PowerEdge R660 with Intel Xeon Gold 5416S | 10x speed up - processed in 1 hour |
DELL PowerEdge R760 with 2x Intel Xeon Platinum 8470 | 20x speed up - processed in 30 minutes |