Image requirements for the Neural tracker and its sub-detectors

Hardware requirements for neural analytics operation

Requirements for the operation of the Neural tracker and its sub-detectors are given in the table:

Camera
  • The optimal resolution for solving typical tasks is 1280x720. We don't recommend using a resolution higher than 1920×1080, since a higher resolution doesn't increase the detection quality but significantly increases the consumption of resources. The minimum resolutions for each neural network are listed in the table:
    Name of the neural networkNeural network (Detector Pack 3.16)Neural network (Detector Pack 3.15)Minimum resolution
    VehicleGeneralNMCar_v1.0_onnxGeneralNMCar_v1.0CPU_IR10_FP16300x300
    GeneralNMCar_v1.0GPU_onnx
    PersonGeneralNMHuman_v1.0_onnxGeneralNMHuman_v1.0_IR10_FP16
    GeneralNMHuman_v1.0GPU_onnx
    Person (top-down view)GeneralNMHumanTopView_Nano_v1.0_onnxGeneralNMHumanTopView_v0.8CPU_IR10_FP16
    GeneralNMHumanTopView_v0.8GPU_onnx
    Person (top-down view Nano)
    		GeneralNMHumanTopView_Nano_v1.0_onnxGeneralNMHumanTopView_Nano_v1.0_CPU_IR10_FP16576x320
    GeneralNMHumanTopView_Nano_v1.0_GPU_onnx
    Person (top-down view Medium)GeneralNMHumanTopView_Medium_v1.0_onnxGeneralNMHumanTopView_Medium_v1.0_CPU_IR10_FP16
    GeneralNMHumanTopView_Medium_v1.0_GPU_onnx
    Person (top-down view Large)
    		GeneralNMHumanTopView_Large_v1.0_onnxGeneralNMHumanTopView_Large_v1.0_CPU_IR10_FP16
    GeneralNMHumanTopView_Large_v1.0_GPU_onnx
    Person and vehicle (Nano)GeneralNMHumanAndVehicle_Nano_v1.0_onnxGeneralNMHumanAndVehicle_Nano_v1.0_CPU_IR10_FP161152x640
    GeneralNMHumanAndVehicle_Nano_v1.0_GPU_onnx
    Person and vehicle (Medium)GeneralNMHumanAndVehicle_Medium_v1.0_onnxGeneralNMHumanAndVehicle_Medium_v1.0_CPU_IR10_FP16
    GeneralNMHumanAndVehicle_Medium_v1.0_GPU_onnx
    Person and vehicle (Large)GeneralNMHumanAndVehicle_Large_v1.0_onnxGeneralNMHumanAndVehicle_Large_v1.0_CPU_IR10_FP16
    GeneralNMHumanAndVehicle_Large_v1.0_GPU_onnx

  • The frame rate per second in the video stream from the camera is at least 8 for solving typical tasks
  • Neural tracker-based analytics works with both colorless (gray) and color images. When you use a neural filter, we recommend using color cameras. When you use black-and-white cameras, the recognition quality can be noticeably worse
  • Correct operation of the detector isn't guaranteed if you use frames that were masked before they got into Axxon One by an IP camera or similar methods.
  • Correct operation of the Neural tracker isn't guaranteed if you use a fisheye camera.
Lighting
  • Lighting in the scene is at least 50 lux. In conditions of insufficient or excessive lighting (night or light-striking), stable operation of the analytics isn't guaranteed
  • No abrupt changes in lighting
Scene and camera angle
  • Moving objects are visually separable from each other
  • The background is mostly static and doesn't change abruptly
  • Moving objects are minimally obscured by static objects in the scene (columns, trees, and so on)
  • The analyzed scene doesn't have reflective surfaces and sharp shadows from moving objects. If present, they must be masked
  • Camera shake doesn't result in image offsets greater than 1% of the frame size