Dockerized human detection API with FastAPI and Detectron2

This modification to detectron2 exposes an API (using FastAPI) which for each picture "batch" creates a .json (format specified by project client) if a person was detected in the picture, and how confident detectron2 is with that label.

Setup assumed a Windows 10/11, with a working Ubuntu 20.04 on WSL2, and Docker Desktop (set with WSL2 backend)

Steps to run

1. clone this repo on your device

git clone https://github.com/v4k0nd/designprojectm11.git

2. Start a shell and enter

cd designprojectm11/docker

3. Build the image (might need to sudo),
   Dockerfile.folder for the folder input output version

   docker build --build-arg USER_ID=$UID -t detectron2:folder-v0 -f Dockerfile.folder .

   Dockerfile.server for the server API call version

   docker build --build-arg USER_ID=$UID -t detectron2:server-v0 -f Dockerfile.server .

4. Run it (might need to sudo) for the folder version

docker run -p 9976:9976 --gpus all -it \
    -v /detectron2_detection/:/code/ \
    -v $(pwd)/inputs:/home/appuser/detectron2_repo/inputs \
    -v $(pwd)/outputs:/home/appuser/detectron2_repo/outputs \
    --env INPUT_DIR=/home/appuser/detectron2_repo/inputs \
    --env OUTPUT_DIR=/home/appuser/detectron2_repo/outputs \
    --name=detectron2_container_folder detectron2:folder-v0

the server version

docker run -p 9976:9976 --gpus all -it -v /detectron2_detection/:/code/ --name=detectron2_container_folder detectron2:server-v0

5. You should be in the docker container, and FastAPI server should have started

6. Open another shell, create dataset-1-5 in main directory, and put some pictures in it

7. To test it working, run

python3 script/test_api.py