Project SGIL

The goal of this project is to use CV paired with satellite imagery to improve localization of a cheap GPS.

Setup

This project is for Linux systems. If on Windows, please use WSL. Also, depending on the images you're using, you will need to download them separately.

Setup Environment

We use uv for our project. You can install it here

# Setup venv with uv
uv sync --all-extras

Lint and format the code

uv run ruff format .
uv run ruff check . --fix

Run the code

# While in the root of the project, run:
uv run python -m project_sgil.manual_sgil

Manual tree labeling

Use the manual labeling tool to click tree centers on the Raleigh satellite image. Clicks are saved to a CSV next to the image and existing points are reloaded when you reopen the tool.

The CSV format is two columns: lat,lon with no header.

# Run the labeling tool
uv run python scripts\manual_sat_tree_detection.py

Convert old pixel CSV to lat/lon

If you have an older pixel-based CSV, run:

uv run python scripts\convert_tree_csv_to_latlon.py

uv run python -m project_sgil.automated_sgil | Out-File -FilePath "dataset\tables\automated_sgil_table.txt" -Encoding utf8

This overwrites RaleighSatellite_manual_trees.csv with lat/lon values and writes a backup at RaleighSatellite_manual_trees.csv.bak.

uv run python .\scripts\grid_search_automated_sgil_configs.py