Blitz API

Blitz API is a RESTful service for the Blitz application.

Requirements

• Minimum of 16 GiB RAM and a CPU with 10 cores
• Docker

Technologies used

• Python v3.11
• Poetry v1.7.1
• Celery (Task Queue)
• Redis (Broker and Datastore for Celery)
• PIFuHD
• lightweight-human-pose-estimation

API Documentation

Api documentation can be found in the following endpoint of the application once deployed.

/docs

Run Application from the pre-built image

Install docker and docker-compose, ignore this step if already installed.

sudo apt install docker docker-compose -y

Clone the github repository.

git clone https://github.com/SpoooKyHammer/Blitz-API.git

After cloning, cd into the directory and the first thing you should do is install the pre-trained model that our project depends on. (Note these commands are to be executed from the root directory of the project).

• To download the PIFuHD model run the following command.
  

cd blitz_api/ext/tasks/models/pifuhd/ && sh scripts/download_trained_model.sh

• To download the human_pose_estimation model run the following command.

cd blitz_api/ext/tasks/models/human_pose_estimation/ && wget https://download.01.org/opencv/openvino_training_extensions/models/human_pose_estimation/checkpoint_iter_370000.pth

Docker login to the remote registry where our image is currently located

sudo docker login anicreate.azurecr.io

Pull the image and run the containers

sudo docker-compose -f docker-compose.prod.yml up -d

Run Application by locally building the image

Install docker and docker-compose, ignore this step if already installed.

sudo apt install docker docker-compose -y

Clone the github repository.

git clone https://github.com/SpoooKyHammer/Blitz-API.git

After cloning, cd into the directory and the first thing you should do is install the pre-trained model that our project depends on. (Note these commands are to be executed from the root directory of the project).

• To download the PIFuHD model run the following command.
  

cd blitz_api/ext/tasks/models/pifuhd/ && sh scripts/download_trained_model.sh

• To download the human_pose_estimation model run the following command.

cd blitz_api/ext/tasks/models/human_pose_estimation/ && wget https://download.01.org/opencv/openvino_training_extensions/models/human_pose_estimation/checkpoint_iter_370000.pth

We will use docker-compose to automatically build the image and run the container.

sudo docker-compose up -d