Advanced Settings#

Configuration Options#

Local Image Building#

By default, the application uses pre-built Docker images for faster setup. If you need to build images locally (for customization or development):

# Build and run locally instead of using pre-built images
make run-lp RENDER_MODE=1 REGISTRY=false

# Apply to any command
make <command> REGISTRY=false

# Examples:
make benchmark REGISTRY=false
make benchmark-stream-density REGISTRY=false

When to use local building:

Modifying source code or configurations
Development and testing changes
Air-gapped environments without internet access
Custom hardware optimizations

Note: Local building takes significantly longer (15-30 minutes) compared to pre-built images (2-5 minutes).

Step-by-Step Local Build Process#

To build the images locally step by step:

Follow the following steps:

make download-models REGISTRY=false
make update-submodules REGISTRY=false
make download-sample-videos
make run-render-mode REGISTRY=false

The above series of commands can be executed using only one command:
```
make run-lp REGISTRY=false RENDER_MODE=1
```

User Defined Workloads#

The application is highly configurable via JSON files in the configs/ directory

To try a new camera or workload:

Create new camera to workload mapping in configs/camera_to_workload_custom.json to add your camera and assign workloads.
- camera_to_workload_custom.json: Maps each camera to one or more workloads. - To add or remove a camera, edit the lane_config.cameras array in the file. - Each camera entry can specify its video source, region of interest, and assigned workloads. Example: json { "lane_config": { "cameras": [ { "camera_id": "cam1", "streamUri": "rtsp://rtsp-streamer:8554/video-stream-name", "workloads": ["items_in_basket", "multi_product_identification"], "region_of_interest": {"x": 100, "y": 100, "x2": 800, "y2": 600} } ] } } If adding new videos, place your video files in the directory performance-tools/sample-media/ and update the streamUri path.

Connecting External RTSP Cameras: To use real RTSP cameras instead of the built-in server:

{
  "camera_id": "external_cam1",
  "streamUri": "rtsp://192.168.1.100:554/stream1",
  "workloads": ["items_in_basket"]
}

Create new configs/workload_to_pipeline_custom.json to define pipeline for your workload.

workload_to_pipeline_custom.json: Maps each workload name to a pipeline definition (sequence of GStreamer elements and models). Example:

{
  "workload_pipeline_map": {
    "custom_workload_1": [
      {
        "type": "gvadetect",
        "model": "yolo11n",
        "precision": "INT8",
        "device": "CPU"
      },
      {
        "type": "gvaclassify",
        "model": "efficientnet-v2-b0",
        "precision": "INT8",
        "device": "CPU"
      }
    ],
    "custom_workload_2": [
      {
        "type": "gvadetect",
        "model": "yolo11n",
        "precision": "INT16",
        "device": "NPU"
      },
      {
        "type": "gvaclassify",
        "model": "efficientnet-v2-b0",
        "precision": "INT16",
        "device": "NPU"
      }
    ],
    "custom_workload_3": [
      {
        "type": "gvadetect",
        "model": "yolo11n",
        "precision": "INT8",
        "device": "GPU"
      },
      {
        "type": "gvaclassify",
        "model": "efficientnet-v2-b0",
        "precision": "INT8",
        "device": "GPU"
      }
    ]
  }
}

Run validate configs command, to verify configuration files

make validate-all-configs

Re-run the pipeline as described above.

[!NOTE] Since the GStreamer pipeline is generated dynamically based on the provided configuration(camera_to_workload and workload_to_pipeline json), the pipeline.sh file gets updated every time the user runs make run-lp or make benchmark. This ensures that the pipeline reflects the latest changes.
src/pipelines/pipeline.sh

Complete Workload Configuration Matrix#

The preconfigured workloads support multiple hardware configurations out of the box. Use the CAMERA_STREAM and WORKLOAD_DIST variables to customize which cameras and hardware (CPU, GPU, NPU) are used by your pipeline.

Usage Pattern:

CAMERA_STREAM=<camera_stream> WORKLOAD_DIST=<workload_dist> make run-lp
# Or for benchmarking:
CAMERA_STREAM=<camera_stream> WORKLOAD_DIST=<workload_dist> make benchmark

Loss Prevention Configurations#

Description	CAMERA_STREAM	WORKLOAD_DIST
CPU (Default)	camera_to_workload.json	workload_to_pipeline.json
GPU	camera_to_workload.json	workload_to_pipeline_gpu.json
NPU + GPU	camera_to_workload.json	workload_to_pipeline_gpu-npu.json
Heterogeneous	camera_to_workload.json	workload_to_pipeline_hetero.json
VLM	camera_to_workload_vlm.json	workload_to_pipeline_vlm.json

[!NOTE] Included Sub-Workloads: The following detection types are automatically enabled in all Loss Prevention configurations:

items_in_basket - Monitors items placed in shopping baskets

hidden_items - Detects concealed or hidden products

fake_scan_detection - Identifies scanning without actual item registration

multi_product_identification - Tracks multiple products simultaneously

product_switching - Detects when customers switch high-value items for lower-value ones

sweet_heartening - Monitors for collusion between customers and cashiers

Automated Self-Checkout Configurations#

Description	CAMERA_STREAM	WORKLOAD_DIST
Object Detection (CPU)	camera_to_workload_asc_object_detection.json	workload_to_pipeline_asc_object_detection_cpu.json
Object Detection (GPU)	camera_to_workload_asc_object_detection.json	workload_to_pipeline_asc_object_detection_gpu.json
Object Detection (NPU)	camera_to_workload_asc_object_detection.json	workload_to_pipeline_asc_object_detection_npu.json
Object Detection & Classification (CPU)	camera_to_workload_asc_object_detection_classification.json	workload_to_pipeline_asc_object_detection_classification_cpu.json
Object Detection & Classification (GPU)	camera_to_workload_asc_object_detection_classification.json	workload_to_pipeline_asc_object_detection_classification_gpu.json
Object Detection & Classification (NPU)	camera_to_workload_asc_object_detection_classification.json	workload_to_pipeline_asc_object_detection_classification_npu.json
Age Prediction & Face Detection (CPU)	camera_to_workload_asc_age_verification.json	workload_to_pipeline_asc_age_verification_cpu.json
Age Prediction & Face Detection (GPU)	camera_to_workload_asc_age_verification.json	workload_to_pipeline_asc_age_verification_gpu.json
Age Prediction & Face Detection (NPU)	camera_to_workload_asc_age_verification.json	workload_to_pipeline_asc_age_verification_npu.json
Age Verification Heterogeneous	camera_to_workload_asc_age_verification.json	workload_to_pipeline_asc_age_verification_hetero.json
Object Detection Heterogeneous	camera_to_workload_asc_object_detection_classification.json	workload_to_pipeline_asc_object_detection_classification_hetero.json