Skip to content
This repository has been archived by the owner on Jul 10, 2023. It is now read-only.

Latest commit

 

History

History
102 lines (78 loc) · 3.63 KB

README.md

File metadata and controls

102 lines (78 loc) · 3.63 KB

Model Downloader

| Specifying Models | Downloading Models | Reference |

The model downloader downloads and prepares models from the OpenVINO™ Toolkit Open Model Zoo for use with Intel(R) Deep Learning Streamer (Intel(R) DL Streamer) Pipeline Server. It can be run as a standalone tool or as part of the Pipeline Server build process. For more information on model file formats and the directory structure used by Intel(R) DL Streamer Pipeline Server see defining_pipelines.

Specifying Models

Models are specified using a yaml file containing a list of model entries. Model entries can either be strings (model names) or objects (model names plus additional properties) and a single yaml file can contain both forms of entries. An example file is used as part of the default build process.

String Entries

String entries specify the model to download from the Open Model Zoo. The model and model-proc files will be downloaded and stored locally using default values.

Example:

- mobilenet-ssd
- emotions-recognition-retail-0003

Object Entries

Object entries specfiy the model to download from the Open Model Zoo and one or more optional properties (alias, version, precision, local model-proc). If an optional property is not specified the downloader will use default values.

Note: Models can have a separate file that contains labels.

Example:

- model: yolo-v3-tf
  alias: object_detection
  version: 2
  precision: [FP32]
  model-proc: object_detection.json
  labels: coco.txt

The model-proc and labels entries above can be set if a local override is desired. In that case, the corresponding files are expected to be in the same directory as the models.list.yml specified.

Default Values

  • alias = model_name
  • version = 1
  • precision = all available precisions
  • model-proc = <model_name>.json
  • labels = .txt

If a local model-proc and/or labels file(s) are not specified, the tool will use the model-proc and/or labels file that is part of the Intel(R) DL Streamer developer image.

Downloading Models

The model downloader can be run either as a standalone tool or as part of the Intel(R) DL Streamer Pipeline Server build process.

Downloading Models as part of Intel(R) DL Streamer Pipeline Server Build

The Intel(R) DL Streamer Pipeline Server build script downloads models listed in a yaml file that can be specified via the --models argument.

Example:

./docker/build.sh --models models_list/models.list.yml

Downloading Models with the standalone tool

When run as a standalone tool, the model downloader will run within an openvino/ubuntu20_data_dev:2021.4.2 docker image and download models listed in a yaml file that can be specified via the --model-list argument.

Example:

mkdir standalone_models
./tools/model_downloader/model_downloader.sh --model-list models_list/models.list.yml --output ${PWD}/standalone_models

Command Line Reference

usage: model_downloader.sh
  [--output absolute path where to save models]
  [--model-list input file with model names and properties]
  [--force force download and conversion of existing models]
  [--open-model-zoo-version specify the version of OpenVINO(TM) image to be used for downloading models from Open Model Zoo]
  [--dry-run print commands without executing]