About the tracker plugin improvement advise

[471747996]

Please provide complete information as applicable to your setup.

**• Hardware Platform (Jetson / GPU) 1068

• DeepStream Version 6.0

• JetPack Version (valid for Jetson only)

• TensorRT Version 8.0

• NVIDIA GPU Driver Version (valid for GPU only) 510

• Issue Type( questions, new requirements, bugs) new requirements

• Requirement details( This is for new requirement. Including the module name-for which plugin or for which sample application, the function description)

nvstracker plugin

I advise this nvtracker plugin need to add two addition property.

one is named like enable_save_tracking_history, which means if enable this plugin to save the history tracking list for each current tracking object, true is enabled and false is disabled.

the other one is named like max_history_locations_to_be_saved, which means max length of the history tracking list for

a tracked object, add this property is in consideration of the memery usage.

This Requirement is not the same as enable-past-frame (which store the current not reported low confidence tracking history),

I advise you to add these to property for the nvtracker plugin, is because the the hisory tracking list for each object

is very important in the video analysis applications,in especial

like application of behavior analysis and event analysis, we need the object’s history tracking info to do

the analysis and then make the decision if one event/behavior is happened on this object.

current nvstracker’s tracking only output all of the current tracking objects’s--------curent frame tracking id,

current frame tracking box and so on, which is all tracking info of a object is current, there is not any

history infos of the current tracking objects.

last I advise to attach the object’s tracking history list to the NvDsObjectMeta, so user can convenient to access

this data.

I have created one topic about this Requirement, but has no any Valuable reply.

Tracking history trajectorys

我测试了deepstream5.1中的跟踪器，当我 g_object_set (G_OBJECT (nvtracker), “enable-past-frame”, 1, NULL); 后，通过probe解析NvDsPastFrameObjBatch时，获取不到历史轨迹数据，三种跟踪算法都试了，都不行。 历史轨迹对于视频分析很重要，我提议如下： 1：将历史轨迹数据附着到每个obj的meta数据里，而不是另外弄一个NvDsPastFrameObjBatch，这样同一个目标的元数据很分散，不利于分析使用。 2：添加两个配属性，like：max_history_locations(代表最大保留的跟踪轨迹数量），enable_save_history_locations(是否启用）。

[kesong]

Thanks for your advices. Can you share your use case?

[471747996]

for example，we want to detect the car’s traffic event，like car stop event happened on the no parking road, we could use the car object’s latest history tracking lcations(for example latest 3 seconds)to anlysis if this event happen<if the car’s last 3 second’s location range of activities is very small ，we could make a decision may be this car is stoped>
the other example we want to detect the car’s converse traffic event, we also need to to use the history tracking list to analyais<for example the normal driving direction the tracking y coordinate should be increased，so if we analysis the car’s history tracking list and find its y coordinate is decreasing，then we could make a judge that this car is driving in converse direction and this event should be reported.

and we also could use the tracking history list with the time stamp to compute the car’s speed

and so on.

Just kidding, I think I need to join you to make deepstream better serve video analysis applications