Quick and dirty cobbled object detector using yad2k and putting it in a python flask application.
The purpose of this app is to give me a quick example object detection service I can deploy and demonstrate with other applications.
There are a ton of nasty dependencies so it is probably best to use the docker image. The downside is that the container image is about 2GB.
Run it with:
docker run -d -p 5005:5005 vallard/photo-detect
Or on a Kubernetes Cluster:
kubectl apply -f https://raw.githubusercontent.com/vallard/YOLO-Detector/master/manifests/yolo-detector.yaml
run:
curl -H "Content-Type: application/json" \
-X GET \
-d '{"url" : "http://www.havayolu101.com/wp-content/uploads/2016/04/Qantas_Boeing-737_aircraft_Tesla_Model-S-P90D_electric-car_001.jpg"}' \
localhost:5005/detect
run from a pod inside the cluster (or expose the service to a LoadBalancer):
curl -H "Content-Type: application/json" \
-X GET \
-d '{"url" : "http://www.havayolu101.com/wp-content/uploads/2016/04/Qantas_Boeing-737_aircraft_Tesla_Model-S-P90D_electric-car_001.jpg"}' \
yolo:5005/detect
This will give us an output of:
[{"item": "aeroplane", "score": "0.68712324"}, {"item": "car", "score": "0.935317"}]If you want to run it without kubernetes or docker then you can look at the Dockerfile to see the dependencies.
This can detect the following inside of an image:
- aeroplane
- bicycle
- bird
- boat
- bottle
- bus
- car
- cat
- chair
- cow
- diningtable
- dog
- horse
- motorbike
- person
- pottedplant
- sheep
- sofa
- train
- tvmonitor
The bulk of the code as mentioned is from yad2k.
There will probably not be too many updates to this code as its just a quick project I did but contributions and suggestions are more than welcome.