KPI
Accuracy
Perception Pipeline
Retail Object (Binary) Detection: Retail Object Detection v1.0 accuracy
Retail Object Embedding Generation: Retail Object Recognition v1.0 accuracy
Performance
Perception Pipeline
Runtime profiling of the Perception (DeepStream) pipeline. T4 supports up to 4 streams and A30 up to 12 streams at 30 FPS.
GPU |
Number of Streams |
PGIE Batch Size |
SGIE Batch Size |
Max FPS |
CPU Utilization (%) |
GPU Utilization (%) |
|---|---|---|---|---|---|---|
T4 |
1 |
1 |
16 |
110 |
2.6 |
100 |
T4 |
4 |
4 |
4 |
30 |
2.9 |
100 |
T4 |
4 |
4 |
16 |
32 |
2.8 |
100 |
T4 |
8 |
8 |
16 |
15 |
2.6 |
100 |
A30 |
1 |
1 |
16 |
310 |
1.3 |
99 |
A30 |
8 |
8 |
16 |
45 |
1.2 |
100 |
A30 |
12 |
12 |
16 |
32 |
1.3 |
100 |
Model Inference Resolution:
Retail Object (Binary) Detection v1.0: FP16 inference
Retail Object Recognition v1.0: FP16 inference
System Configuration:
NVIDIA T4 GPU : Intel(R) Xeon(R) Silver 4114 CPU @ 2.20 GHz
NVIDIA A30 GPU: Intel(R) Xeon(R) Gold 6338N CPU @ 2.20 GHz
Media Microservices
Runtime profiling of the media microservices:
Components |
GPU Utilization |
CPU Memory |
CPU Cores Allocated |
CPU Utilization |
Comment |
|---|---|---|---|---|---|
VST |
4% |
35.0 GiB |
2 |
91% |
GPU usage will go up based on how many users are accessing the UI and using the overlay feature. |
Number of Input RTSP Streams: 12
System Configuration:
GPU: NVIDIA T4
CPU: Intel(R) Xeon(R) Gold 6126 @ 2.60 GHz