Overview

This is an object detection project which uses the YOLO object detection algorithm. The object detection algorithm is exposed via FastAPI

Running

Make sure you have Docker installed.

In order to run this project you will need to download 3 missing files:

1. coco.names - https://raw.githubusercontent.com/pjreddie/darknet/master/data/coco.names - the names of the classes for the model
2. yolov3.cfg - https://github.com/pjreddie/darknet/blob/master/cfg/yolov3.cfg - config file for the model - can get this here .
3. yolov3.weights - https://pjreddie.com/media/files/yolov3.weights - weights for the model.

Once these are downloaded, place them in an artifacts/config folder - ypu should have a folder structure like the following:

├── README.md
├── artifacts
│   ├── config
│   │   ├── coco.names
│   │   ├── yolov3.cfg
│   │   └── yolov3.weights
│   └── outputs
├── docker-compose.yml
└── src
    ├── Dockerfile
    ├── __init__.py
    ├── main.py
    ├── obj_detection
    │   ├── __init__.py
    │   ├── object_detection.py
    │   └── routes.py
    └── requirements.txt

To run the project type the following to build the image:

docker-compoose build

Once the image is built, run the following to run the project:

docker-compose up

The artifacts folder is mounted in the container and this is where files processed will be saved.

Also, all code in the src folder is mounted to the container so any changes will be seen in realtime.

Once running, head to http://0.0.0.0:80/docs to test out the API

This is all setup to be run in local development mode, if you want to build for production, remoove the following line from the docker-compose.yml file:

command: /start-reload.sh

Other

The following lines are included in the Dockerfile as the openCV library was unable to load in the inage.

RUN apt-get update ##[edited]
RUN apt-get install 'ffmpeg'\
    'libsm6'\ 
    'libxext6'  -y

The foloowng volumes are mounted in docker-compose.yml:

volumes: 
    - "/Users/martin/projects/object_detection/artifacts/config/:${CONFIG_DIR}"
    - "/Users/martin/projects/object_detection/artifacts/outputs/:${ARTIFACT_DIR}"
    - /Users/martin/projects/object_detection/src:/src/

The first 2 mounts are for the config files (for the yolov3 model) and an output directory for processed images/videos. These are setup on my local machine and need to be managed by me - in order to industrialize this type of system it would be better to use a bind mount instead with some 3rd party provider (AWS/Azure/Google cloud). See https://docs.docker.com/storage/volumes/ for more info.

is for development purposes only, this should be replaced with a proper bind mount in production.