VCAserver Hardware Recommendations
Minimum Hardware Recommendations:
| Optimal Camera Configuration:
|
Note: VCA will accept higher resolutions and frame rates but they will increase the CPU overhead and result in a reduction in channel capacity without necessarily increasing accuracy.
All performance benchmarks were performed against VCAserver v2.1.0.
*Applies to all license types
Ubuntu 18.04
Tested CPUs Recommend using the top model range from each series | Tracker/Classifier | |||||||
Object Tracker | DL Filter | DL Object Tracker | DL People Tracker | DL Skeleton Tracker (appropriate GPU required) | Hand Object Interaction (appropriate GPU required) | Behaviour (DLST, Fall, Aggressive) (appropriate GPU required) | Forensics (DLOT + REID) (appropriate GPU required) | |
Intel i3-13100F | 75 | 75 | 25 | 25 | 25 | 25 | 7 | 12 |
Intel i5-13600 | 120 | 110 | 80 | 80 | 75 | 70 | 40 | 40 |
Intel i7-13700KF | 230 | 220 | 115 | 115 | 90 | 110 | 55 | 70 |
Jetson Orin Nano 4GB (ABP 301) | 4 | 4 | 4 | 4 | 2 | N/A | ||
Jetson Orin Nano 8GB (ABP 302) | 7 | 5 | 5 | 5 | 3 | 4 | ||
Intel Xeon 4208 | ||||||||
Intel Xeon 4210 | ||||||||
Intel Xeon 4216 | ||||||||
Intel Xeon 4314 | 250 | 240 | 100 | 100 | 85 | 95 | 50 | 70 |
Ubuntu 18.04
Tested NVIDIA GPUs | DL Feature | |||||
DL Filter | DL Object Tracker | DL People Tracker | DL Skeleton Tracker (appropriate GPU required) | Hand Object Interaction (appropriate GPU required) | Behaviour (DLST, Fall, Aggressive) (appropriate GPU required) | |
GTX 1660 Super | 75 | 25 | 25 | 25 | 25 | 7 |
RTX 4060 | 80 | 80 | 75 | 70 | 40 | |
RTX 4060ti | 220 | 115 | 115 | 90 | 110 | 55 |
A2000 | 200 | 35 | 35 | 25 | 30 | 25 |
A4000 | 150 | 60 | 60 | 40 | 55 | 40 |
A5000 | 250 | 90 | 90 | 70 | 65 | 50 |
Feature | Description | License Required |
Deep Learning Filter (DLF) | Provides classification for people or vehicles detected by VCA which have passed through defined zones and rule | VCApresenceAi, VCAproAi |
Deep Learning Object Tracker (DLOT) | [1]Used for the accurate classification and tracking of people, vehicles and select objects. | VCAproAi |
Deep Learning People Tracker (DLPT) | [2]Used for the accurate classification and tracking of people, usually in a well-lit area. | VCAproAi |
Deep Learning Skeleton Tracker (DLST) | Used for the definition of skeletal keypoints and tracking of people, usually in a well-lit area. | VCAproAi, VCAbehaviour |
Hand Object Interaction (HOI) Tracker | Used for the detection and tracking of hands and objects held in hands, from a top down field of view | VCAbehaviour |
Behaviour | Channel configuration including DLST, fall detection and aggressive behaviour analytics | VCAbehaviour |
Analytic metadata is available from all channels and is processed in real-time, metadata can be received via SSE or RTSP metadata streams. The volume of metadata produced will vary based on the number of objects tracked and rules configured for a given channel. As a guide, a video stream with an average of 10 objects tracked will generate between 420 and 550 kbps of metadata.
[1]Deep Learning Object Tracker works exclusively with the NVIDIA CUDA supported range of GPUs
[2]Deep Learning People Tracker works exclusively with the NVIDIA CUDA supported range of GPUs