Deploy Mounted Files

Dayu system requires deploying files to support mounting of component containers.

Download files from mounted files link, it has the following structure:

dayu-deploy-files/
|-- cloud-deploy-files/
|   `-- dayu-files/
|       |-- acc-gt-dense/
|       |   |-- car-detection/
|       |   `-- face-detection/
|       |-- acc-gt-sparse/
|       |   |-- car-detection/
|       |   `-- face-detection/
|       |-- processor/
|       |   |-- age-classification/
|       |   |-- car-detection/
|       |   |-- category-identification/
|       |   |-- exposure-identification/
|       |   |-- face-detection/
|       |   |-- gender-classification/
|       |   |-- license-plate-recognition/
|       |   |-- model-switch-detection/
|       |   |-- pedestrian-detection/
|       |   `-- vehicle-detection/
|       `-- scheduler/
|           `-- fixed/
|-- datasource-deploy-files/
|   `-- datasource/
|       `-- video/
|           |-- road_0/
|           |-- road_1/
|           |-- road_dense/
|           |-- road_sparse/
|           |-- street_0/
|           |-- street_dense/
|           `-- street_sparse/
|-- edge-deploy-files/
|   `-- dayu-files/
|       |-- generator/
|       |   `-- adaptive/
|       `-- processor/
|           |-- age-classification/
|           |-- car-detection/
|           |-- category-identification/
|           |-- exposure-identification/
|           |-- face-detection/
|           |-- gender-classification/
|           |-- license-plate-recognition/
|           |-- model-switch-detection/
|           |-- pedestrian-detection/
|           `-- vehicle-detection/
`-- tensorrt-application-files/
    |-- age-classification/
    |-- car-detection/
    |-- category-identification/
    |-- exposure-identification/
    |-- face-detection/
    |-- gender-classification/
    |-- license-plate-recognition/
    |-- pedestrian-detection/
    `-- vehicle-detection/

Deploy to Nodes

The default mount paths are configured in template/base.yaml:

default-file-mount-prefix: "/data/dayu-files"
datasource:
  data-root: "/data/datasource/"

If you keep the default values, deploy the downloaded files as follows:

Downloaded directory	Target node	Target path	Usage
cloud-deploy-files/dayu-files/	Cloud node	/data/dayu-files/	Files mounted by cloud-side processors and schedulers.
edge-deploy-files/dayu-files/	Each edge node that may run Dayu components	/data/dayu-files/	Files mounted by edge-side processors and generators.
datasource-deploy-files/datasource/	Edge node configured by datasource.node	/data/datasource/	Video files used by the simulated datasource when it is enabled.
tensorrt-application-files/<application>/	Optional source files for cloud and edge nodes	See below	Platform-specific TensorRT engines used when replacing model files.

For example, with the default paths, processor/car-detection/ in a processor template is resolved to /data/dayu-files/processor/car-detection/ on the node where the processor runs. Make sure that the corresponding directory exists on both cloud and edge nodes if a template uses position: both.

TensorRT Application Files

cloud-deploy-files and edge-deploy-files already contain model files that match the default cloud and edge deployment. tensorrt-application-files is provided for replacing or preparing TensorRT engine files for different hardware platforms. Each application directory is grouped by platform:

tensorrt-application-files/<application>/
|-- cloud/
|-- jp4-tx2/
|-- jp4-xavier-nx/
|-- jp5-agx-xavier/
`-- jp6-orin-nano/

Select the platform directory that matches the target node, then copy the engine files into the same application directory under /data/dayu-files/processor/. For example, the cloud-side files for car-detection should be placed in /data/dayu-files/processor/car-detection/ on the cloud node, while the JetPack 4 TX2 files should be placed in the same relative path on the corresponding TX2 edge node.

If you change default-file-mount-prefix or datasource.data-root in template/base.yaml, deploy the files to the new paths instead of the default paths shown above.