Docker container for OpenVINO Model Zoo tools (omz_downloader, omz_converter, omz_data_downloader, omz_info_dumper)

This project provides a Dockerfile to build a Docker image to run Intel's Open Model Zoo support tools easily.

Building image and preparation

cd Docker
docker build -t omz:2024.6.0 .
docker tag omz:2024.6.0 omz:latest
cd ..
makdir -p models
# To make the Docker container user (openvino) is able to write the files in the directory
chmod 777 models

Usage - Downloading and converting model

The downloaded models will be found in the ./models directory after running following commands.

docker run --rm -v ./models:/home/openvino/models omz omz_downloader --name yolo-v3-onnx
docker run --rm -v ./models:/home/openvino/models omz omz_converter --name yolo-v3-onnx --precisions FP16

Windows users need to specify a full path for the model directory.

docker run --rm -v D:/<foo>/Docker/models:/home/openvino/models omz omz_downloader --name yolo-v3-onnx

Usage - Display the names of all supported models

You can check which models are supported by the tool with following command.

docker run --rm -v ./models:/home/openvino/models omz omz_downloader --print_all

Usage - Work in the container console

You can work in the container terminal to do multiple steps.

docker run --rm -it -v ./models:/home/openvino/models omz

Usage - Download and convert all available models

docker run --rm -v ./models:/home/openvino/models omz omz_downloader --all
docker run --rm -v ./models:/home/openvino/models omz omz_converter --all --precisions FP16

Reference

• OpenVINO Open Model Zoo (OMZ)
• List of available models at OMZ
  • Intel pre-trained models
  • Public pre-trained models
• OMZ downloader tools

You can use NNCF to compress/quantize the DL models.

• Intel Neural Network Compression Framework (NNCF)