DwarfView is a Python-based tool designed to query and visualize cutouts of dwarf galaxy candidates detected in the UNIONS surveys. It retrieves image cutouts based on celestial coordinates (Right Ascension and Declination) or a catalog of objects and generates plots in either a grid or channel-by-channel format. It supports custom input catalog files (CSV or Parquet), single-coordinate queries, and configuration via a YAML file or command-line arguments.
- Input Options: Query cutouts using a single RA/Dec pair, a custom catalog (CSV or Parquet), or the default unions_master.parquet catalog.
- Filtering: Filter candidates by prediction probability (zoobot_pred) and optionally exclude known objects or training data.
- Visualization: Plot cutouts in a grid layout (RGB images) or as individual channels (R, G, B) plus RGB composite.
- Configurability: Adjust settings via a config.yaml file or override them with command-line arguments.
- CADC account
- Experience with the CANFAR science platform
Install required packaged using pip
pip install -r requirements.txt- Clone the repository
git clone https://github.com/heesters-nick/DwarfView.git
cd DwarfView- Install the package to make dwarfview available as a command-line tool
pip install .- Update main_dir, data_dir, etc., in config.jaml to match your system.
DwarfView can be run with default settings or customized via CLI arguments or the config.yaml file.
Visualize 50 candidates from the unions dwarf catalog catalog with predicted probabilities p_zoobot > 0.8:
dwarfviewPlot cutout for a specific object
dwarfview --ra 123.4567 --dec 89.1011Use a catalog of objects to query. CSV and Parquet file are currently supported (must contain 'ra' and 'dec' columns):
dwarfview --input-file path/to/catalog.parquetFilter candidates with probabilities between 0.5 and 0.9:
dwarfview --p-min 0.5 --p-max 0.9Display individual channels plus RGB image instead of a grid plot of RGB cutouts:
dwarfview --plot-mode channelSave the figure to a specific file:
dwarfview --output-file my_cutouts.png