KPI

Accuracy

Perception Pipeline

  • People Detection: PeopleNet v2.6 accuracy

  • PeopleNet Transformer: (FAN-S + DINO) accuracy

  • ReidentificationNet v1.1: accuracy

  • ReIdentificationNet Transformer (Swin Tiny): accuracy

Performance

Perception Microservice

Runtime profiling of the Perception (DeepStream) microservice:

Perception Microservice Profiling

Detection Model

ReID Model

GPU

Number of Streams (30 FPS)

CPU Utilization

GPU Utilization

PeopleNet v2.6 (ResNet-34)

ReIdentificationNet v1.1

NVIDIA T4

12

3.2%

95%

PeopleNet v2.6 (ResNet-34)

ReIdentificationNet v1.1

NVIDIA A30

23

1.5%

93%

PeopleNet v2.6 (ResNet-34)

ReIdentificationNet v1.1

NVIDIA A100

27

380%

80%

PeopleNet Transformer

ReIdentificationNet Transformer (Swin Tiny)

NVIDIA A100

3

147%

86%

Model Inference Resolution:

  • PeopleNet v2.6: INT8 inference

  • ReidentificationNet v1.1: FP16 inference

  • ReIdentificationNet Transformer: FP16 inference

  • PeopleNet Transformer: FP16 inference

Metrics obtained on the below system configuration:

  • NVIDIA T4 GPU : Intel(R) Xeon(R) Gold 6126 CPU @ 2.60 GHz

  • NVIDIA A30 GPU: Intel(R) Xeon(R) Gold 6338N CPU @ 2.20 GHz

  • NVIDIA A100 GPU: Intel(R) Xeon(R) Gold 6258R CPU @ 2.70GHz

Analytics Microservices

Runtime profiling of the analytics & API components:

Each one is profiled separately. Perception metadata is from a perception microservice, on a different machine but in the same cluster, processing RTSP streams.

Profiling of Analytics Components

Components

GPU Utilization

CPU Memory

CPU Utilization

Comment

Analytics

N/A

2.344 GiB

90.4%

Behavior Learning - Ingestion

N/A

2.338 GiB

158.25%

Behavior Learning - Training

30%

3.57 GiB

108%

GPU is used only during clustering & model training.

Triton Inference server

10%

615.7 MiB

0.34%

Web API

N/A

81.21 MiB

1%

CPU usage will go up based on how many users accessing the UI.

Kafka

N/A

1.504 GiB

9%

Logstash

N/A

1.103 GiB

23%

Elastic

N/A

1.802 GiB

81%

Number of Input Metadata Streams: 2

Metrics obtained on the below system configuration:

  • GPU: NVIDIA A100

  • CPU: Intel(R) Xeon(R) Gold 6258R CPU @ 2.70GHz

Measurements collected using the following models:

  • ReIdentificationNet Transformer: FP16 inference

  • PeopleNet Transformer: FP16 inference

Media Microservices

Runtime profiling of the media microservices:

Profiling of Analytics Components

Components

GPU Utilization

CPU Memory

CPU Utilization

Comment

VST

Decoder: 25%, Encoder: 25%, GPU: 15%

27.36 GiB

53.07%

Max streams: 30. GPU usage will go up based on how many users are accessing the UI and using the overlay feature.

Number of Input RTSP Streams: 30

Metrics obtained on the below system configuration:

  • GPU: NVIDIA A100

  • CPU: Intel(R) Xeon(R) Gold 6258R CPU @ 2.70GHz