Install¶

IDFace Match Server is provided as a Docker image via Amazon ECR. The current version of the server and the docker pull command for it are in your personal account on the customer portal.

Install Docker¶

Since the IDFace Match Server is distributed in a form of Docker image, you need to make sure that you have Docker Engine installed on your machine. Follow the Docker's instruction to install Docker Engine on Ubuntu, CentOS or other supported platform. For Windows you can use Docker Desktop.

Note that by default you need to run docker with sudo. You can disable it if you want.

Download image¶

Follow these steps to install an image via the custom portal:

1. Navigate to your personal account on the custom portal and select the desired image.
2. In the Docker containers section, you can choose between CPU and GPU builds.
3. Click the "Generate command" button for your selected build.
4. Wait for the command to be generated. Once ready, click the "Copy" button to copy the pull command.
5. Use the copied command in Terminal or another tool.

For the CPU build, the image name will be:

561397722827.dkr.ecr.eu-west-1.amazonaws.com/growthproject/facerecognition/idfacematch/idfacematch-server-eval:#.##.#

For the GPU build, the image name will be:

561397722827.dkr.ecr.eu-west-1.amazonaws.com/growthproject/facerecognition/idfacematch/idfacematch-server-eval:#.##.#-gpu

Once the Docker image is pulled, you may want to change its name since the original name is quite long. Use docker tag to create a second short image name:

For CPU version:

docker tag \
    561397722827.dkr.ecr.eu-west-1.amazonaws.com/growthproject/facerecognition/idfacematch/idfacematch-server-eval:#.##.# \
    idfacematch-server-eval:#.##.#

For GPU version:

docker tag \
    561397722827.dkr.ecr.eu-west-1.amazonaws.com/growthproject/facerecognition/idfacematch/idfacematch-server-eval:#.##.#-gpu \
    idfacematch-server-eval:#.##.#-gpu

Now you can use the shorter image name with the Docker commands. The old name is still available if you need it.

GPU Support¶

IDFace Match Server is available in both CPU and GPU versions. The GPU version offers significantly improved performance for face detection and template generation operations.

GPU Requirements¶

To run the GPU version of IDFace Match Server, you need:

• NVIDIA GPU with compute capability at least 7.5
• NVIDIA driver that supports CUDA 12
• NVIDIA Container Toolkit installed on the host machine

Note that if you have CUDA or TensorRT already installed on your host system, they won't be used by the container. The server uses its own libraries bundled within the Docker image.