galaxy.segmentation

Script to create segmentation maps for models.

To create segmentation maps for randomly chosen clusters, random objects and stars use function saveSegMaps(). To create a segmentation map with larger scale for a randomly chosen cluster use function saveBigSegMap().

Functions

cleanup_memory(→ None)	Cleans up memory by deleting variables and freeing GPU memory.
load_model(→ torch.nn.Module)	Loads a trained model with specific weights.
create_sample(→ tuple[pandas.DataFrame, ...)	Ensures the segmentation sample exists by downloading or generating it.
grab_surrounding(→ Generator[tuple[int, int], None, None])	Generates surrounding grid points for a given radius.
ddos_segmentation_map(map_type, ra_start, dec_start, ...)
create_map_dataloader(→ torch.utils.data.DataLoader)	Creates a dataloader for a segmentation map.
prepare_sample_dataloaders(→ list[tuple[int, ...)	Prepares dataloaders for segmentation map samples.
create_segmentation_plot(→ None)	Creates a segmentation map plot for a given sample.
create_segmentation_plots(→ None)	Creates segmentation maps for all samples.

Module Contents

galaxy.segmentation.cleanup_memory(*args) → None

Cleans up memory by deleting variables and freeing GPU memory.

Args:

*args: Variables to be deleted.

galaxy.segmentation.load_model(model: torch.nn.Module, n_in_channels: int, optimizer_name: str, device: torch.device) → torch.nn.Module

Loads a trained model with specific weights.

Args:

model (torch.nn.Module): The model to be loaded. optimizer_name (str): Name of the optimizer used during training. device (torch.device): Device to load the model on.

Returns:

torch.nn.Module: The loaded model.

galaxy.segmentation.create_sample(sample_name: galaxy.util.SampleName2, dataset: str) → tuple[pandas.DataFrame, galaxy.util.MapType]

Ensures the segmentation sample exists by downloading or generating it.

Returns:

tuple: Sample dataframe and map type.

galaxy.segmentation.grab_surrounding(points_on_radius: int) → Generator[tuple[int, int], None, None]

Generates surrounding grid points for a given radius.

Args:

points_on_radius (int): Radius in grid points.

Yields:

tuple[int, int]: Coordinates of grid points.

galaxy.segmentation.ddos_segmentation_map(map_type: galaxy.util.MapType, ra_start: float, dec_start: float, map_dir: pathlib.Path, description_path: pathlib.Path, already_ddosed=False)

galaxy.segmentation.create_map_dataloader(map_dir: pathlib.Path, dataset: str) → torch.utils.data.DataLoader

Creates a dataloader for a segmentation map.

Args:

map_type (MapType): Type of the map. ra_start (float): Starting RA coordinate. dec_start (float): Starting Dec coordinate. map_dir (Path): Directory to save the map data. survey (str, optional): Survey name. Defaults to None. bands (list, optional): List of bands. Defaults to None. imgsize_pix (int, optional): Image size in pixels. Defaults to 224.

Returns:

DataLoader: Dataloader object for the segmentation map.

galaxy.segmentation.prepare_sample_dataloaders(data: pandas.DataFrame, sample_name: galaxy.util.SampleName2, dataset: str, map_type: galaxy.util.MapType) → list[tuple[int, torch.utils.data.DataLoader]]

Prepares dataloaders for segmentation map samples.

Args:

data (pd.DataFrame): Data describing the samples. sample_name (SampleName2): Name of the sample. map_type (MapType): Type of map to generate.

Returns:

list[tuple[int, DataLoader]]: List of tuples containing index and dataloader.

galaxy.segmentation.create_segmentation_plot(model_name: str, optimizer_name: str, num_epochs: int, predictor: galaxy.train.Predictor, dataset: str, sample_name: galaxy.util.SampleName2, n_cols: int = 5) → None

Creates a segmentation map plot for a given sample.

Args:

model_name (str): Name of the model. optimizer_name (str): Name of the optimizer used during training. predictor (train.Predictor): Predictor for making predictions. sample_name (SampleName2): Name of the sample. n_cols (int, optional): Number of columns in the plot. Defaults to 5.

galaxy.segmentation.create_segmentation_plots(model: torch.nn.Module, model_name: str, dataset: str, n_in_channels: int, num_epochs: int, optimizer_name: str, map_type: galaxy.util.MapType = MapType.SMALL) → None

Creates segmentation maps for all samples.

Args:

model (torch.nn.Module): The model used for predictions. model_name (str): Name of the model. optimizer_name (str): Name of the optimizer used during training. map_type (MapType, optional): Type of map to generate. Defaults to MapType.SMALL.

Returns:

None