A web tool for performing object detection in agricultural images.
The AgriCounter tool is divided into two main components: a Node.js application and a Python process. These two processes run in parallel and communicate using a shared file tree and by sending and receiving http requests.
The site directory in this repository contains the Node.js application.
The backend directory contains the code for the Python process. The Python process is responsible for performing model-related tasks (training, fine-tuning, and prediction).
Follow these steps to create a new Docker instance of AgriCounter.
The agricounter_ctl.py script (located in the root directory of the repository) can be used to create and manage the AgriCounter tool inside a Docker container. In order to set up the application, agricounter_ctl.py will attempt to read a configuration file called args.json. The args.json file should be located in the root directory of the AgriCounter repository.
Included in this repository is an args-template.json file with the required configuration keys. Before creating a Docker instance of AgriCounter, this file needs to be edited with the desired configuration values and renamed to args.json.
Below is an explanation of the keys that agricounter_ctl.py expects to find in the args.json file:
url_prefix: All URL paths will begin with this string. Example: "/agricounter"site_port: The port number that the Node.js application will listen on.backend_python_port: The port number that the Python server process will listen on.postgres_db_name: The name of the PostGres database that AgriCounter will use to store user accounts.postgres_db_username: The username for the PostGres role that has database privileges.postgres_db_password: The password for the PostGres role that has database privileges.postgres_db_port: The port number that the PostGres database will listen on.api_key: The Python process sends this password whenever it sends an HTTP request to the Node.js application. The Node.js application uses this password to verify that the request was sent by the Python process.admin_username: Username for the site's administrator account.admin_password: Password for the site's administrator account.gpu_index: Index of GPU device to use. If only one GPU is available, this should be 0. Use -1 if you want to use the CPU instead.use_slurm: Use SLURM job scheduler (for HPC environments). Requires setup of aslurm_config.jsonfile inbackend/src.
Before creating a Docker instance of AgriCounter, it is also necessary to generate a PEM encoded SSL certificate and private key. A self-signed certificate can be created with the following commands:
cd site/myapp
openssl genrsa -out key.pem
openssl req -new -key key.pem -out csr.pem
openssl x509 -req -days 2000 -in csr.pem -signkey key.pem -out cert.pem
To create the AgriCounter application for the first time, run ./agricounter_ctl.py --create.
To remove the Docker containers without removing the PostGreSQL volume, use ./agricounter_ctl.py --down. The containers can then be rebuilt with ./agricounter_ctl.py --up. To remove the containers and remove the PostGreSQL volume, use ./agricounter_ctl.py --destroy.
When an AgriCounter Docker instance is created for the first time with ./agricounter_ctl.py --create, only the administrator account is seeded to the database. In order to add regular user accounts, it is necessary to log in as the administrator and add the new user accounts through the administrator web page interface. It is also necessary to specify the object classes that the AgriCounter tool will be used to detect (e.g., "Canola Seedling"). This is also done from the administrator's web page.
To set up the AgriCounter tool without Docker, the two main components of the project (located in the site and backend directories) need to be set up independently. Each of these directories contains a README.md file with instructions on how to perform the setup. Once the setup has been performed, the two processes should be started in two separate terminal sessions.