def use_img_classifier_in_mem(
clf, segmenter_args, label_to_colors_args, img_path, out_img
):
img = skimage.io.imread(img_path)
seg, clf = image_segmentation.trainable_segmentation(img, clf=clf, **segmenter_args)
color_seg = shapes_to_segmentations.label_to_colors(seg, **label_to_colors_args)
segimgpil = plot_common.img_array_to_pil_image(color_seg)
segimgpil.save(out_img)
def use_img_classifier_in_mem(
clf, segmenter_args, label_to_colors_args, img_path, out_img
):
img = skimage.io.imread(img_path)
features = multiscale_basic_features(
img,
sigma_min=segmenter_args["sigma_min"],
sigma_max=segmenter_args["sigma_max"],
**segmenter_args["segmentation_features_dict"],
)
seg = predict_segmenter(features, clf)
color_seg = shapes_to_segmentations.label_to_colors(seg, **label_to_colors_args)
segimgpil = plot_common.img_array_to_pil_image(color_seg)
segimgpil.save(out_img)
def show_segmentation(image_path, mask_shapes, features, segmenter_args):
""" adds an image showing segmentations to a figure's layout """
# add 1 because classifier takes 0 to mean no mask
shape_layers = [color_to_class(shape["line"]["color"]) + 1 for shape in mask_shapes]
label_to_colors_args = {
"colormap": class_label_colormap,
"color_class_offset": -1,
}
segimg, _, clf = compute_segmentations(
mask_shapes,
img_path=image_path,
shape_layers=shape_layers,
label_to_colors_args=label_to_colors_args,
features=features,
)
# get the classifier that we can later store in the Store
classifier = save_img_classifier(clf, label_to_colors_args, segmenter_args)
segimgpng = plot_common.img_array_to_pil_image(segimg)
return (segimgpng, classifier)