YOLO Object Detection Service

Dockerized object detection service using YOLO based on AlexeyAB's darknet fork and exposed as a REST API using connexion.

Quick start

You can build and run with:

docker-compose up -d --build; docker-compose logs -f

docker will expose the service over localhost:80 and you can see a OpenAPI Specification at localhost:80/ui

Env Vars

$DOCKER_REPO should be a valid docker image such as aquintero446/yolo_service

Installation

Build the image

docker build -t $DOCKER_REPO:1.0-yolov3_coco \
	-f ./Dockerfile.dev \
	--build-arg weights_file="yolov3.weights" \
	--build-arg config_file="data/yolov3.cfg" \
	--build-arg meta_file="data/coco.data" \
       	.

Follow vars will change the YOLO model:

weights_file
config_file
meta_file

Dependencies

Automatically installed when using the steps above.

Usage

Docker: docker run -dit --rm --name test_yolo -p 9999:8080 $DOCKER_REPO:1.0-yolov3_coco

Docker compose: docker-compose up

Send request:

curl -X POST -F 'image_file=@person.jpg' -F threshold=0.25 'http://localhost:9999/detect'
curl -X GET 'http://localhost:9999/detect?url=https%3A%2F%2Fgithub.com%2FAlexeyAB%2Fdarknet%2Fraw%2Fmaster%2Fdata%2Fperson.jpg'

Travis

You must configure some env variables to work properly:

DOCKER_ID docker user/id
DOCKER_PASSWORD docker password
DOCKER_REPO docker repo with docker user id userid/tpkutils-server
AWS_ACCESS_KEY aws access key credential
AWS_SECRET_KEY aws access secret credential