Automating Fireballs

Repository of work for Desert Fireball Network research. Contains code for automated point pickings and YOLO fireball detection.

Installation

Clone repository.

git clone https://github.com/Snaiel/Automating-Fireballs.git

Enter directory.

cd Automating-Fireballs/

Create virtual environment.

python3 -m venv .vev

Activate virtual environment.

source .venv/bin/activate

Install dependencies.

pip install -r requirements.txt

Directories

point_pickings contains code for automating the point pickings process.

object_detection contains code for training and testing a YOLO object detection model for fireballs in tiles.

fireball_detection contains code for tiling and performing detections on full-sized images.

Usage

Everything is designed to work (hopefully) from the project directory. If a particular script/file is meant to be run directly, you run it as a module. For example, to run

point_pickings/misc/scikit_blobs.py

use the following command:

python3 -m point_pickings.misc.scikit_blobs

Notice how . is used for package separators and .py is omitted. Tab completion isn't available which does suck when typing things out...

Fireball Detection

fireball_detection.detect has the detection system implemented as a function where you call detect_fireballs with an image and it returns bounding boxes and confidence of detected fireballs.

Running the module also shows a sample detection.

python3 -m fireball_detection.detect

Fireball Point Pickings

point_pickings.process has the automated point picking system implemented as a function where you call retrieve_fireball with an image (assumed to be cropped) and returns all the processed position and timing information.

Running the module shows a sample with the distances between segments being labelled.

python3 -m point_pickings.process

Running point_pickings.compare shows a comparison between the automated system and the manual point pickings.

python3 -m point_pickings.compare