synthval.features_extraction.FeatureExtractor#

class synthval.features_extraction.FeatureExtractor#

Bases: abc.ABC

Abstract base class representing a generic feature extractor. Child classes must implement the batch_feature_extraction method for batch processing.

abstractmethod batch_feature_extraction(images)#

Abstract method to extract features from a batch of images.

Parameters:

images (List[PIL.Image.Image]) – List of PIL images to extract features from.

Returns:

A NumPy array containing extracted features for the batch.

Return type:

numpy.ndarray

group_feature_extraction(source_folder_path, batch_size=16, verbose=True)#

Extract features from all images in a specified folder using batch processing.

Parameters:

  • source_folder_path (str) – Path to the folder containing images.

  • batch_size (int, optional) – Number of images to process in a single batch (default: 16).

  • verbose (bool, optional) – If True, logs progress during extraction (default: True).

Returns:

A DataFrame containing extracted features for all images.

Return type:

pandas.DataFrame

get_features_df(source_folder_path, save_path=None, batch_size=16, verbose=True)#

Extract features from images in a folder and optionally save them to a CSV file.

Parameters:

  • source_folder_path (str) – Path to the folder containing images.

  • save_path (str, optional) – Path to save the features DataFrame as a CSV file. If a CSV file exists, it will be loaded instead.

  • batch_size (int, optional) – Number of images to process per batch (default: 16).

  • verbose (bool, optional) – If True, logs progress during extraction (default: True).

Returns:

DataFrame containing extracted features for all images.

Return type:

pandas.DataFrame