SYNAPSIS
Tutorials  /  Deep dive · Installation
Deep dive Script drafted ~5 min Companion to 02 · MaskAnyone

Installation & setup

Docker, GPU support, starting the MaskAnyone stack, and recognising a healthy install.

Video coming soon
Script drafted · Recording scheduled in batch 2 production (≈ 2027)

Learning goal

After watching, you can install Docker, start the MaskAnyone stack, and reach the web interface on your own machine. You know how to recognise a successful install and where to look when things break.

Transcript

Cold open

Five minutes from now, you'll have MaskAnyone running on your own machine. Let's go.

Check your hardware

Two checkpoints before you start. Minimum: 16 GB RAM, modern multi-core CPU, 50 GB free disk. Recommended for serious work: 32 GB RAM, an NVIDIA GPU with 8 GB or more of VRAM, an SSD. CPU-only works — it's just slow.

No GPU at home? You have options. Cloud compute by the hour. Your university's HPC. Or just team up with someone who has the hardware. Later tutorial.

Install Docker

Step one — install Docker Desktop. Go to docker.com/products/docker-desktop, download for your operating system, run the installer.

On Windows, enable WSL 2 when prompted, then restart. On Mac, drag Docker into Applications and grant the permissions it asks for. On Linux, install via your package manager.

Verify with the version check and a test container:

docker --version
docker run hello-world

If you see "Hello from Docker", you're done with step one.

GPU support — optional

If you don't have an NVIDIA GPU, skip this section. Everyone else, this is where you wire your GPU into Docker.

On Linux, four commands install the NVIDIA Container Toolkit and configure Docker to use it. On Windows with WSL 2, GPU access is automatic if your NVIDIA drivers are current.

docker run --rm --gpus all nvidia/cuda:11.0-base nvidia-smi

If that command returns a GPU table, you're set.

Start MaskAnyone

Step two — clone the infrastructure repo. One git command, one cd, one docker-compose up -d. That's it.

git clone https://github.com/MaskAnyone/MaskAnyoneProdInfrastructure.git
cd MaskAnyoneProdInfrastructure
docker-compose up -d

Heads-up: first start downloads roughly 15 GB of machine-learning models. Plan for 10 to 30 minutes depending on your internet. After this, future starts take seconds.

Once it's done, docker-compose ps should show five containers up and healthy: frontend, backend, sam2, openpose, db. If any say "restarting" or "exited", you have a problem — see the troubleshooting deep-dive.

Open the interface

Open localhost:8080 in your browser. You should see the MaskAnyone dashboard. Upload area in the middle, project list below. You're done.

Recap

Three things to remember. One: Docker is the foundation — install it first, run hello-world, prove it works. Two: docker-compose up -d from the MaskAnyone infrastructure repo brings the whole stack up. Three: first run is slow, every run after is fast.

If anything failed — port collision, memory error, GPU not detected — the troubleshooting deep-dive walks through every common installation problem we've seen.