Test task : You need to write a server to classify the picture then containerize it in the docker and make the server globally accessible.

Note : app was developed on ubuntu 20.04. Docker was tested on ubuntu and macos. On macOs it can be warnings messages .

⚙️ Setup & Launch

Note : Download libtorch before build local app

1) clone repo 
git clone https://github.com/vladimirlabynko/image_classification.git

2) go to repo directory
cd image_classification 

3) make build folder and build it 
mkdir build
cmake .. or  cmake .. -DCMAKE_PREFIX_PATH=path/to/libtorch..
make

4) After building app run app
./image classification

After app start you'll see next message :

== Model [../resnet18.pt] loaded!
== Label loaded! Let's try it

While server is running,open another terminal and send CURL requests from your local pc like this :

curl -X POST -d "urltoimage" http://localhost:12345

Wait a seconds and you'll see a predicted class like this :

Class: mask

Docker

To run docker you'll need next steps in terminal: 1)Clone docker :

docker push 5nevil/image_classification:v2

2)Run docker , for example with command :

docker run --rm -it -p 8020:12345 5nevil/image_classification:v2

3. Open another terminal and make POST curl request :

curl -X POST -d "urltoimage" http://localhost:8020

Wait a seconds and you'll see a predicted class like this :

Class: mask

If you want to classify local files you'll need next steps in terminal:

1)Clone docker :

docker push 5nevil/image_classification:v1

2)Run docker , for example with command :

docker run --rm -it -p 8020:12345  -v /path/to/local/images:/images 5nevil/image_classification:v1

3. Then when you in docker run commands :

cd build 
./image_classification

4. Open another terminal and write next comaand :

curl -X POST -d "/images/yorfile.jpg(png...)" http://0.0.0.0:8020