Computer Vision Initialization Script

Note: Running with real models are not tested yet. (18/04/2022)

Prerequisites

• python >= 3.8.5 (For running main.py)
  • python 3.6 should be fine, although it is tested with python 3.8.5
• conda as environment manager
• pyyaml library

Installation

You should install each model as described on each model's README.md

• Object Detection: https://github.com/robocup-eic/robocup2022-cv-object-detection
• What Is That: https://github.com/robocup-eic/robocup2022-cv-what-is-that
• Person Tracker: https://github.com/robocup-eic/robocup2022-cv-person-tracker

Each of the model's libraries must be configured through conda environment from requirements.txt provided in each repository.

# Create new conda env
conda create --name "<env_name>" python="<version>"
# Activating newly created env
conda activate "<env_name>"
# Install lib using conda
conda install --file "<requirements.txt>"
# Install lib using pip
pip install -r "<requirements.txt>"
# Coming back to base env
conda deactivate

Take note of each model's file path and environment name, you will have to edit the config.yaml using these settings.

Configuration

Edit the config.yaml with each model's main script file path and the conda environment name.

model1:
  conda_env: "<model1_env_name>"
  exec_path: "<main_file_to_execute.py>"
model2:
  conda_env: "<model2_env_name>"
  exec_path: "<main_file_to_execute.py>"

Each model must be provided with conda_env and exec_path. The exec_path can be either absolute or relative. Multiple models (or any process/program actually) can be appended to the config.yaml as shown in the above example.

Executing

Start: Run python script as normal

python main.py

Stop: Type CTRL+C in the terminal console

Socket API Servers

• Object Detection: tcp://localhost:10001
• What Is That: tcp://localhost:10000
• Person Tracker: tcp://localhost:11000

For API Syntax, please refer to each repository.