Tracker Information

Hi Guys,

I am using tracker in my application. I am trying to figure out which objects are moving in this frame. Is there a way to get this information from tracker? What is the theory behind the tracker? Can the tracker handle disappearances or does the tracking only happen when the object is detected in successive frames?

Thanks.

Could you refer https://docs.nvidia.com/metropolis/deepstream/4.0/DeepStream_Plugin_Manual.pdf#%5B%7B%22num%22%3A56%2C%22gen%22%3A0%7D%2C%7B%22name%22%3A%22XYZ%22%7D%2C106%2C223%2C0%5D section 2.2 firstly, the low level lib are KLT/IOU/DCF, you can select one using config.

Hi bcao,

Thanks for the pointer. I would like to extend the use of tracker to a longer time. As understood in the webinar as well as the other posts, I altered the interval parameter to a non-zero value in the file ‘config_infer_primary_nano.txt’ as I am working on Nano. When I make the changes, I do not see any difference in fps as well as detections. Ideally, the fps should increase proportionally with interval parameter. Is this the right place to make the change?

[property]
gpu-id=0
net-scale-factor=0.0039215697906911373
model-file=../../models/Primary_Detector_Nano/resnet10.caffemodel
proto-file=../../models/Primary_Detector_Nano/resnet10.prototxt
model-engine-file=../../models/Primary_Detector_Nano/resnet10.caffemodel_b8_fp16.engine
labelfile-path=../../models/Primary_Detector_Nano/labels.txt
batch-size=8
process-mode=1
model-color-format=0
## 0=FP32, 1=INT8, 2=FP16 mode
network-mode=2
num-detected-classes=4
<b>interval=10</b>
gie-unique-id=1
output-blob-names=conv2d_bbox;conv2d_cov/Sigmoid
#parse-bbox-func-name=NvDsInferParseCustomResnet
#custom-lib-path=/path/to/libnvdsparsebbox.so
#enable-dbscan=1

Thanks

There’s a deepstream-app config file like source1.txt as a sample. You would find a section called [primary-gie] in the config file. Under that section, you can find interval=0, where you can change it to a non-zero value. I remember this intervaloverwrites a default PGIE config params shown in your reply. So, please change interval value in deepstream-app config file, rather than PGIE config file.

Regarding the performance change, it depends on where is the perf bottleneck. If the CPU is the bottleneck (likely when KLT tracker is used), then changing the PGIE interval wouldn’t have much impact on the perf. If the bottleneck was in the GPU, then yes, you would get perf gain by increasing inference interval. You can check CPU and GPU utilization % in Nano while running your pipeline to identify what is the bottleneck.