Clone the repository with the command git clone --recursive -j8 [url], because it contains submodules used in the repo
Or if the repo was already cloned, run git submodule update --init
Before running the scripts, make sure you have the following dependencies installed in the requirements.txt
or you can run poetry install to install them using pypoetry
After this, run poetry shell to create an environment that will contain all the deps you need for the execution of the scripts
In your computer should be installed Docker.
In the root of the project, type the following terminal command:
sudo docker-compose up -d
In the folder il_5safe/gui there is a folder called frontend
Inside that you can find a web application, realized in VueJS (v2)
In the docker-compose.yml there are two principal containers, one that manage the backend (check Dockerfile_backend in the root path), a simple container that executes a Flask python script that creates a route for being called by the frontend (check Dockerfile_frontend).
The docker-compose file pulls the images that are available in the dockerhub storage. If something is wrong or not working, it is recommended to comment the rows 4 and 20 (the images directives), and build again everything on your own computer.
Be sure to download and place the model trained with our custom dataset in the folder il_5safe/resources/weights/yolov5/best_model.pt
before the build
Execute the command sudo docker-compose build --no-cache.
For the multiplatform build: docker buildx build --platform linux/amd64,linux/arm64 -f Dockerfile_backend --push --tag giorginogreg/innovationlab-5safe-be:latest .
Then execute the command sudo docker-compose up -d
After the build / docker-compose up command is executed successfully, you can open the frontend interface in the browser via the link http://localhost:8080
Then, the interface will explain you the steps to execute, for example uploading an image and then clicking the prediction button after the model selection
Test command: /opt/poetry-venv/bin/python models/yolov5/detect.py --weights resources/weights/yolov5/best_model.pt --source uploaded_image.jpg --name uploaded_image --exist-ok