DensePose API Docker

GPU powered Docker API that recieves an image and returns the DensePose (old Detectron version) UVmask:

Input

Output

Dependencies

• Any Linux distribution and a GPU (RTX GPUs not working yet)

• Docker

• Docker-Compose

• nvidia-Docker

Usage

1. From the docker folder run: docker-compose up -d

After building the Docker container (it can take a while), the API will start.

You can check the status of the container with: docker ps -a

You can also see what's going on inside the container with docker logs densepose-api -f

2. On the main folder run: python send_files.py

If it's the first time you send an image to the API it can take a while as it will download the DensePose model (607MB)

This will send demo_im.jpg to the API using the Python requests package.

Feel free to modify the script and add it to your pipeline

Tested on Ubuntu 18.04 with a GTX 1060 GPU (drivers 410.104)

Advantages

• You don't have to install Caffe

• You don't have to install CUDA

• Very easy to integrate in any pipeline, only a HTTP request needed.

• The API will be isolated from your code. No dependency errors

• Did I mention that you don't need to install Caffe?