relai package

Subpackages

• relai.configurations package
  • Module contents
• relai.datasets package
  • Submodules
  • relai.datasets.coco_format module
    • CocoFormatDataset
  • relai.datasets.image_folder module
    • ImageFolderDataset
      • ImageFolderDataset.class_name_to_images
      • ImageFolderDataset.classes
    • navigate_to_root()
  • relai.datasets.relai_dataset module
    • RELAIClassificationDataset
      • RELAIClassificationDataset.class_idx_to_images
      • RELAIClassificationDataset.class_name_to_images
      • RELAIClassificationDataset.classes
      • RELAIClassificationDataset.num_classes
    • RELAIDataFormat
      • RELAIDataFormat.COCOFORMAT
      • RELAIDataFormat.IMAGEFOLDER
    • RELAIDataset
    • Subset
      • Subset.class_name_to_images
    • TransformDataset
  • relai.datasets.segmentation_dataset module
    • RELAISegmentationDataset
      • RELAISegmentationDataset.get_dataset()
      • RELAISegmentationDataset.load_dataset()
    • RELAISegmentationDatasetCSV
      • RELAISegmentationDatasetCSV.get_dataset()
      • RELAISegmentationDatasetCSV.load_dataset()
    • SegmentationDataPair
      • SegmentationDataPair.ground_truth
      • SegmentationDataPair.ground_truth_path
      • SegmentationDataPair.input_image
      • SegmentationDataPair.input_image_path
  • relai.datasets.utils module
    • coco_collate()
    • get_dataset()
  • Module contents
• relai.metrics package
  • Submodules
  • relai.metrics.metrics module
    • Confidence
      • Confidence.reduction
      • Confidence.forward()
      • Confidence.forward()
    • Hits
      • Hits.reduction
      • Hits.forward()
      • Hits.forward()
    • RELAIMetric
      • RELAIMetric.forward()
    • dice()
    • dice_segmentation()
    • false_negative_segmentation()
    • false_positive_segmentation()
    • iou()
    • iou_segmentation()
    • precision()
    • precision_segmentation()
    • true_negative_segmentation()
    • true_positive_segmentation()
  • relai.metrics.statistics module
    • compute_significance()
    • compute_significance2()
  • Module contents
• relai.models package
  • Submodules
  • relai.models.cifar10_cnn module
    • CIFAR10_CNN
      • CIFAR10_CNN.backbone
      • CIFAR10_CNN.default_transform
      • CIFAR10_CNN.forward()
      • CIFAR10_CNN.last_fc
  • relai.models.relai_model module
    • RELAIModel
      • RELAIModel.backbone
      • RELAIModel.default_transform
      • RELAIModel.finetune()
      • RELAIModel.last_fc
      • RELAIModel.load_model_state()
      • RELAIModel.model_state
  • relai.models.resnet module
    • ResNet50
      • ResNet50.backbone
      • ResNet50.default_transform
      • ResNet50.forward()
      • ResNet50.last_fc
  • relai.models.robust_resnet module
    • RobustResNet50
      • RobustResNet50.backbone
      • RobustResNet50.default_transform
      • RobustResNet50.forward()
      • RobustResNet50.forward_features()
      • RobustResNet50.last_fc
    • load_pretrained_robust_model()
    • load_robust_model()
    • make_and_restore_robust_model()
  • relai.models.timm_model module
    • TimmModel
      • TimmModel.default_transform
      • TimmModel.forward()
      • TimmModel.get_normalizer()
      • TimmModel.last_fc
      • TimmModel.load_model_state()
      • TimmModel.model_state
  • relai.models.timm_vit module
    • TimmVit
      • TimmVit.backbone
      • TimmVit.default_transform
      • TimmVit.forward()
      • TimmVit.last_fc
  • relai.models.utils module
    • get_detection_model()
    • get_model()
    • modify_user_checkpoint_keys()
    • validate_checkpoint()
  • Module contents
• relai.text package
  • Subpackages
    • LLM Analysis
      • relai.text.llm.inspect package
  • Module contents
• relai.transforms package
  • Submodules
  • relai.transforms.transforms module
    • HStack
    • Identity
    • Zero
  • Module contents
• relai.utils package
  • Submodules
  • relai.utils.download module
    • download_relai_resource()
  • relai.utils.file_handling module
    • save_imgs_in_dir()
    • zip_directory()
  • relai.utils.image_handling module
    • convert_to_pil()
    • fig2img()
  • Module contents
• relai.vision package
  • Subpackages
    • Image Classification
      • Inspections
      • Mitigations
      • Datasets
    • relai.vision.data_utils package
      • Submodules
      • relai.vision.data_utils.filter module
      • relai.vision.data_utils.image_sorter module
      • relai.vision.data_utils.image_tagger module
      • relai.vision.data_utils.sort_dataset module
      • relai.vision.data_utils.tag_dataset module
      • Module contents
    • Object Detection
      • Inspections
      • Datasets
    • Image Segmentation
      • Inspections
      • Datasets
  • Submodules
  • relai.vision.robustness_test module
    • CorruptionWrapper
    • RGB2BGR
    • RobustnessTest
      • RobustnessTest.run()
      • RobustnessTest.summarize()
    • brightness()
    • clipped_zoom()
    • contrast()
    • defocus_blur()
    • disk()
    • fog()
    • gaussian_noise()
    • get_average_precision()
    • get_test_type_description()
    • impulse_noise()
    • motion_blur()
    • rain()
    • robustness_analysis()
    • shadow()
    • shot_noise()
    • snow()
    • sun_flare()
    • versatile_analysis()
    • zoom_blur()
  • Module contents

Submodules

relai.relai_algorithm module

class relai.relai_algorithm.RELAIAlgorithm(fabric: Fabric = None)

Bases: object

Base class for all algorithms in the relai_algorithms package. Any users of the relai_algorithms package should not need to interact with this class directly.

• Parameters:
  • fabric (L.Fabric , optional) – Lightning fabric to run the algorithm. If None, a new fabric with 1 device will be created. Defaults to None.
  • save_to_path (Union *[*str , Path ] , optional) – Path to save the outputs to. If None, the outputs will be saved to the current working directory with a default name that includes the algorithm name and a timestamp. If provided and the parent directories do not exist, they will be silently created. Defaults to None.

relai.relai_params_validator module

class relai.relai_params_validator.RELAIParamsValidator(schema)

Bases: object

property errors : list

validate(params: dict)

Module contents