def get_slides()

in sampling.py [0:0]

• 69 lines of code
• 14 McCabe index (conditional complexity)

def get_slides(database: Database,
               test_slide_ids: List[int],
               base_path: str = 'WSI',
               size: int = 256,
               positive_class: int = 2,
               negative_class: int = 7):
    
    lbl_bbox = []
    training_slides = []
    test_slides = []
    files = []
    
    slides = tqdm(
        database.execute("SELECT uid, filename FROM Slides").fetchall(),
        desc='Loading slides...',
    )

    for idx, (cur_slide, filename) in enumerate(slides):

        database.loadIntoMemory(cur_slide)
        slide_path = os.path.join(base_path, filename)
        slide = openslide.open_slide(slide_path)

        level = 0
        level_dimension = slide.level_dimensions[level]
        down_factor = slide.level_downsamples[level]

        # Dictionary of classes found so far, initialized with the positive class, which
        # we are sure it exists.
        classes = {
            positive_class: 1
        }

        annotations = {}
        labels, bboxes = [], []
        for id, annotation in database.annotations.items():

            # Ignore deleted annotations, or any that are not of type SPOT.
            if annotation.deleted or annotation.annotationType != AnnotationType.SPOT:
                continue

            annotation.r = 25
            d = 2 * annotation.r / down_factor
            x_min = (annotation.x1 - annotation.r) / down_factor
            y_min = (annotation.y1 - annotation.r) / down_factor
            x_max = x_min + d
            y_max = y_min + d

            # If the annotation class is not yet in the classes dictionary, create
            # an entry and initialize the structures within.
            if annotation.agreedClass not in annotations:
                annotations[annotation.agreedClass] = {}
                annotations[annotation.agreedClass]['bboxes'] = []
                annotations[annotation.agreedClass]['label'] = []

            # Store bounding box and label into the annotation dictionary.
            annotations[annotation.agreedClass]['bboxes'].append([int(x_min), int(y_min), int(x_max), int(y_max)])
            annotations[annotation.agreedClass]['label'].append(annotation.agreedClass)

            # Store the bounding box and its label separately in the bboxes and labels
            # lists.
            if annotation.agreedClass in classes:
                label = classes[annotation.agreedClass]

                #                     bboxes.append([int(x_min), int(y_min), int(x_max), int(y_max)])
                bboxes.append(BoundingBox(int(x_min), int(y_min), int(x_max), int(y_max)))
                labels.append(label)

        if len(bboxes) > 0:
            lbl_bbox.append([bboxes, labels])
            
            is_test = str(cur_slide) in test_slide_ids
            if is_test:
                files.append(SlideContainer(
                    file=slide_path,
                    annotations=annotations,
                    level=level,
                    width=size,
                    height=size,
                    y=[bboxes, labels],
                    sample_func=partial(sampling_func, negative_class=negative_class)))
                test_slides.append(len(files) - 1)
            else:
                files.append(SlideContainer(
                    file=slide_path,
                    annotations=annotations,
                    level=level,
                    width=size,
                    height=size,
                    y=[bboxes, labels],
                    sample_func=partial(sampling_func, negative_class=negative_class)))
                training_slides.append(len(files) - 1)

    return lbl_bbox, training_slides, test_slides, files