Video requirements for scene analytics detection tools

Video stream and scene requirements for Neurotracker operation

Objects image requirements for Neurotracker

Hardware requirements for neural analytics operation

To configure the Scene Analytics detection tools based on Neurotracker, do the following:

  1. Go to the Detection Tools tab.

  2. Below the required camera, click Create… → Category: Trackers → Neurotracker.

By default, the detection tool is enabled and set to detect people.

If necessary, you can change the settings of the detection tool parameters given in the table:

ParameterValueDescription
Object features
Record objects trackingYes

By default, metadata are recorded into the database. To disable metadata recording, select the No value

No
Video streamMain stream

If the camera supports multistreaming, select the stream for which detection is needed

Second stream
Other
EnableYesBy default, the detection tool is enabled. To disable, select the No value
No
NameNeurotrackerEnter the detection tool name or leave the default name
Decoder modeAuto

Select a processing resource for decoding video streams. When you select a GPU, a stand-alone graphics card takes priority (when decoding with NVIDIA NVDEC chips). If there is no appropriate GPU, the decoding will use the Intel Quick Sync Video technology. Otherwise, CPU resources will be used for decoding

CPU
GPU
HuaweiNPU
Neurofilter modeCPU

 Select a processing resource for neural network operation (see Hardware requirements for neural analytics operation, General information on configuring detection).

  • We recommend using the GPU. It may take several minutes to launch the algorithm on NVIDIA GPU after you apply the settings. You can use caching to speed up future launches (see Optimizing the operation of neural analytics on GPU).
  • Starting with Detector Pack 3.11, Intel HDDL and Intel NCS aren’t supported.


Nvidia GPU 0
Nvidia GPU 1
Nvidia GPU 2
Nvidia GPU 3
Intel NCS (not supported)
Intel HDDL (not supported)
Intel GPU
Huawei NPU
Number of frames processed per second6

Specify the number of frames for the neural network to process per second. The higher the value, the more accurate tracking, but the load on the CPU is also higher. The value must be in the range [0.016; 100].

6 FPS or more is recommended. For fast moving objects (running individuals, vehicles), you must set the frame rate at 12 FPS or above (see Examples of configuring Neurotracker for solving typical tasks).

TypeNeurotrackerName of the detection tool type (non-editable field)
Advanced settings
Camera position

Wall To eliminate false positives when using a fisheye camera, select the correct device location. For other devices, this parameter is irrelevant
Ceiling
Hide moving objects

Yes

If you don't need to detect moving objects, select the Yes value. An object is considered static if it doesn't change its position more than 10% of its width or height during its track lifetime

If a static object starts moving, the detection tool will trigger, and the object will no longer be considered static.

No
Hide static objects

Yes

If you don't need to detect static objects, select the Yes value. This parameter lowers the number false positives when detecting moving objects. An object is considered static if it has not moved more than 10% of its width or height during the whole time of its track existence.

If a static object starts moving, the detection tool will trigger, and the object will no longer be considered static.

No
Minimum number of detection triggers6Specify the Minimum number of detection triggers for the neurotracker to display the object's track. The higher the value, the more is the time interval between the detection of an object and the display of its track on the screen. Low values of this parameter may lead to false positives. The value must be in the range [2, 100]
Model quantization

Yes

To quantize the network, select the Yes value. This parameter allows you to reduce the consumption of the GPU processing power.

AxxonSoft conducted a study in which a neural network model was trained to identify the characteristics of the detected object with quantization. The following results of the study were obtained: model quantization can lead to both an increase in the percentage of recognition and a decrease. This is due to the generalization of the mathematical model. The difference in detection ranges within ±1.5%, and the difference in object identification ranges within ±2%.

  • Model quantization is only applicable to NVIDIA GPUs.
  • The first launch of a detection tool with quantization enabled may take longer than a standard launch.
  • If GPU caching is used, next time a detection tool with quantization will run without delay.
No
Neural network file 

If you use a unique neural network, select the corresponding file.

  • To train your neural network, contact AxxonSoft (see Data collection requirements for neural network training).
  • A trained neural network for a particular scene allows you to detect only objects of a certain type (for example, a person, a cyclist, a motorcyclist, and so on).
  • If the neural network file is not specified, the default file will be used, which is selected automatically depending on the selected object type (Object type) and the selected processor for the neural network operation (Decoder mode). If you use a custom neural network, enter a path to the file. The selected object type is ignored when you use a custom neural network.
  • To ensure the correct operation of the neural network on Linux OS, the corresponding file must be located in the /opt/AxxonSoft/DetectorPack/NeuroSDK directory. 
Scanning window

YesTo enable the scanning mode, select the Yes value (see Configuring the Scanning mode)
No
Scanning window height0

The height and width of the scanning window are determined according to the actual size of the frame and the required number of windows. For example, the real frame size is 1920×1080 pixels. To divide the frame into four equal windows, set the width of the scanning window to 960 pixels and the height to 540 pixels

Scanning window step height0

The scanning step determines the relative offset of the windows. If the step is equal to the height and width of the scanning window respectively, the segments will line up one after another. Reducing the height or width of the scanning step will increase the number of windows due to their overlapping each other with an offset. This will increase the detection accuracy, but will also increase the CPU load.

The height and width of the scanning step must not be greater than the height and width of the scanning window—the detection tool will not operate with such settings.

Scanning window step width0
Scanning window width0The height and width of the scanning window are determined according to the actual size of the frame and the required number of windows. For example, the real frame size is 1920×1080 pixels. To divide the frame into four equal windows, set the width of the scanning window to 960 pixels and the height to 540 pixels
Selected object class 

If necessary, specify the class of the detected object. If you want to display tracks of several classes, specify them separated by a comma with a space. For example, 1, 10.
The numerical values of classes for the embedded neural networks: 1—Human/Human (top view), 10—Vehicle.

    1. If you leave the field blank, the tracks of all available classes from the neural network will be displayed (Object type, Neural network file).
    2. If you specify a class/classes from the neural network, the tracks of the specified class/classes will be displayed (Object type, Neural network file).
    3. If you specify a class/classes from the neural network and a class/classes missing from the neural network, the tracks of a class/classes from the neural network will be displayed (Object type, Neural network file).

    4. If you specify a class/classes missing from the neural network, the tracks of all available classes from the neural network will be displayed (Object type, Neural network file).

      Starting with Detector Pack 3.10.2, if you specify a class/classes missing from the neural network, the tracks won’t be displayed (Object type, Neural network file).

Similitude search

Yes

To enable the search for similar persons, select the Yes value. If you enabled the parameter, it increases the processor load.

The Similitude search works only on tracks of people.

No
Time of processing similitude track (sec)0Specify the time in the range [0; 3600] required for the algorithm to process the track to search for similar persons
Time period of excluding static objects0Specify the time in seconds after which the track of the static object is hidden. If the value of the parameter is 0, the track of the static object isn't hidden. The value must be in the range [0; 86 400]
Track retention time0.7Specify the time in seconds after which the object track is considered lost. This helps if objects in scene temporarily overlap each other. For example, a larger vehicle may completely block the smaller one from view. The value must be in the range [0.3, 1000]
Basic settings
Detection threshold30Specify the Detection threshold for objects in percent. If the recognition probability falls below the specified value, the data will be ignored. The higher the value, the higher the accuracy, but some triggers may not be considered. The value must be in the range [0.05, 100]
Neurotracker mode








CPU

Select the processor for the neural network operation (see Hardware requirements for neural analytics operation, General information on configuring detection).

  • We recommend using the GPU. It may take several minutes to launch the algorithm on NVIDIA GPU after you apply the settings. You can use caching to speed up future launches (see Optimizing the operation of neural analytics on GPU).
  • If neurotracker is running on GPU, object tracks may be lagging behind the objects in the Surveillance window. If this happens, set the camera buffer size to 1000 milliseconds (see The Camera object).
  • Starting with Detector Pack 3.11, Intel HDDL and Intel NCS aren’t supported.









Nvidia GPU 0
Nvidia GPU 1
Nvidia GPU 2
Nvidia GPU 3
Intel NCS  (not supported)
Intel HDDL (not supported)
Intel GPU
Huawei NPU
Object type





PersonSelect the recognition object


Person (top-down view)
Vehicle
Person and vehicle (Nano)—low accuracy, low processor load
Person and vehicle (Medium)—medium accuracy, medium processor load
Person and vehicle (Large)—high accuracy, high processor load
Neural network filter
NeurofilterYesTo use the neurofilter to sort out certain tracks, select the Yes value. For example, the neurotracker detects all freight trucks, and the neurofilter sorts out only the tracks that contain trucks with cargo door open
No
Neurofilter file Select a neural network file

If necessary, in the preview window, set detection areas with the help of anchor points  (the same as with the excluded areas of the Scene analytics detection tools, see Setting General Zones for Scene analytics detection tools). By default, the whole frame is a detection area.

For convenience of configuration, you can "freeze" the frame. Click the button. To cancel the action, click this button again.

The detection area is displayed by default. To hide it, click the button. To cancel the action, click this button again.

To save the parameters of the detection tool, click the Apply  button. To cancel the changes, click the Cancel  button.

The next step is to create and configure the necessary detection tools on the basis of neurotracker. The configuration procedure is the same as for the basic tracker (see Setting up Tracker-based Scene Analytics detection tools).

  • To trigger the Motion in Area detection tool on the basis of neurotracker, an object must be displaced by at least 25% of its width or height in FOV.
  • The abandoned objects detection tool works only with the basic object tracker.