Help Center/ Video Intelligent Analysis Service/ Service Overview/ Constraints and Limitations
Updated on 2024-04-03 GMT+08:00
View PDF
Share

Constraints and Limitations

VIAS has certain constraints and limitations, some cost-related and some technological. System-wide constraints affect all sub-services. Additionally, sub-services are subject to their respective constraints.

System-Wide Constraints

  • Supported video formats include AVI, WMV, MPG, MPEG, MP4, MOV, M4V, and MKV.
  • Supported frame rates include 12, 24, 25, and 30 fps.
  • Decoding is supported for H.264 and H.265 videos.

    Encoding Format

    Resolution

    H.264

    720P, 1080P, 2K, 4K

    H.265

    720P, 1080P, 2K, 4K

Constraints of Algorithm Services on Cameras

Table 1 Constraints

Object

Constraint

Camera installation height and angle
  • Camera installation height: 3 m to 5 m
  • Tilted angle: 10° to 40°

Lighting and orientation
  • Adequate lighting at night or in poorly lit indoor environments is required, so that targets are clearly visible in video footage.
  • Avoid direct exposure to strong light.

    After installation is complete, make sure the camera lens is not exposed to harsh light, such as sunlight or street lamps.

  • Avoid light reflections.

    Keep the camera away from highly reflective objects such as glass, ceramic tiles, water, leaves, signboards, and advertisements.

Camera resolution and image quality
  • The camera resolution should be at least 1080p.
  • The video footage must be clear and free from artifacts and overexposure.

Camera installation location
  • The detection area must not have objects (such as trees) blocking the camera's line of sight.
  • The location must meet the installation requirements (for example, poles and walls are available for camera installation).

Constraints and Recommendations on Edge Node Servers

Edge nodes where VIAS is running support Ascend and x86+GP Tnt004 servers.

  • To use Ascend servers, Atlas 800 AI server (model 3000) is recommended.
    • VIAS supports the c82 driver of Ascend inference chip (download path: c82 driver). If the Ascend chip driver of the current server is not of the c82 version, install the c82 driver.
    • On edge node servers, the default ID of the Ascend driver user group HwHiAiUser is 1001. Do not change the value. Otherwise, algorithm jobs may run abnormally. The ID of the HwHiAiUser user group supported by VIAS is 1001, 1000, or 10001.
  • To use x86+GP Tnt004 servers, RH2288 V5 and Atlas G5500 are recommended. CentOS 7.x is recommended for the server operating system. If Tnt004 servers are used, download the recommended Tnt004 driver version.
    Table 2 Recommended server type

    Server Model

    Description

    NPU/GP

    Atlas 800 AI server (model 3000)

    CPU: 2 x Kunpeng 920 5220

    Memory: 256 GB DDR4

    Hard disk: 2 x 480 GB SATA-SSD + RAID 0/1/5

    Atlas 300I Pro: up to eight cards supported

    NIC: 2 x dual-port 25GE NIC (Mellanox CX4)

    Altas 300I Pro

    Atlas G5500

    Operating system: CentOS 7.x/ Ubuntu 18.04/Ubuntu 16.04/RHEL 7.x (CentOS 7.x is recommended.)

    CPU: two Intel® Xeon® SkyLake 6161 v5 CPUs

    Memory: 128 GB

    Hard disk: 2 x 1.2 TB (RAID1) system disk + 1 x 500 GB data disk

    Graphics card (Tnt004): up to eight cards supported (The server supports a maximum of 16 graphics cards. However, only eight graphics cards are supported due to the limitation of CPU cores.)

    NIC: 2 x 10GE

    GP Tnt004

    RH2288H V5

    Operating system: CentOS 7.x/ Ubuntu 16.04/RHEL 7.x (CentOS 7.x is recommended.)

    CPU: two Intel® Xeon® Gold 6154 CPUs

    Memory: 64 GB

    Hard disk: 2 x 600 GB (RAID1) system disk + 1 x 500 GB data disk

    Graphics card (Tnt004): up to two cards supported

    NIC: 2 * GE

    GP Tnt004