2 gvadetect elements in a pipeline, results in decrease of processing FPS

[divdaisymuffin]

Hi @nnshah1,

Recently got a issue with running a pipeline.json , which usage two gvadetect element, It runs fine when I deploy single pod, and gives good processing FPS, if received FPS is 10, process FPS is 9.5.
But when I deploy other pods, FPS keeps on decreasing and the one that deployed last gets the lowest FPS like 2.3, which results in creation of recording snippet with stuck frames.

Pipeline template I am using
"template":"rtspsrc udp-buffer-size=212992 name=source ! queue ! rtph264depay ! h264parse ! video/x-h264 ! tee name=t ! queue ! decodebin ! gvadetect model=\"{models[uniform_yolov4_tiny_608_default_anchors_mask_012_INT8][1][network]}\" model-proc=\"{models[uniform_yolov4_tiny_608_default_anchors_mask_012_INT8][1][proc]}\" name=\"detection1\" ie-config=CPU_THREADS_NUM=2 threshold=0.40 ! gvadetect model=\"{models[head_yolov4_tiny_608_INT8][1][network]}\" model-proc=\"{models[head_yolov4_tiny_608_INT8][1][proc]}\" name=\"detection\" ie-config=CPU_THREADS_NUM=1 threshold=0.40 ! gvametaconvert name=\"metaconvert\" ! gvapython name=\"visitorDensityStaffUniformDet\" module=\"custom_transforms/visitorDensityStaffUniformDet\" class=\"WaitTime\" ! queue ! gvametapublish name=\"destination\" ! appsink name=appsink t. ! splitmuxsink max-size-time=300000000000 name=\"splitmuxsink\"",

Please suggest,the possible solution or reason why this would happen.

[nnshah1]

can you share the system details? Would want to start by determining whether the system can process more than one stream at 10 fps.

[nnshah1]

can you share the CPU utilization for single pod, two pods, etc.

[dhaval-zala-aivid]

CPU configuration ...
Intel(R) Core(TM) i7-8700 CPU @ 3.20GHz
CPU(s): 12
Ram: 24 GB
ssd: 250 GB
hdd: 1 TB

For one, two and three pods CPU utilisation is in between 16% to 18%, 28% and 43% respectively.