Tools

This section outlines the tools and libraries used in the Aerscape Machine Learning project.

We are currently using the following tools:

• Detectron2: A library for object detection and segmentation.
• PyTorch: The primary deep learning framework used for model training and inference.
• OpenCV: Used for image processing tasks.
• VitePress: For generating and hosting project documentation.

Setting up

To start working with this repository, you'll need to set up our imagery and training tools following the instructions below:

1. Create a new Python virtual environment (3.10 ~ 3.12).
2. Install PyTorch, Vision and base libraries:

pip install -r requirements.txt

3. Install Facebook's Detectron2 tool, following the instructions of the official website: Installation. You can also install it on the virtual environment directly like this, but doing so will not let you use the helper scripts from detectron, just the Python APIs.

pip install --no-build-isolation 'git+https://github.com/facebookresearch/detectron2.git'

Using Detectron2

To start familiarizing yourself with Detectron2's API and training options, please read through this section of the documentation:

Getting Started

As well as run through the steps of the following Notebook on Google Colab (it's free) to use, train and test a object detection model:

Detectron2 Tutorial - Collab