import matplotlib as mpl
import matplotlib.pylab as plt
mpl.rcParams["figure.dpi"] == 300
import numpy as np
from skimage.color import label2rgb
from cell_analysis_tools.image_processing import normalize
from cell_analysis_tools.metrics import dice, total_error
[docs]
def image_show(image):
"""
Parameters
----------
image : ndarray
image to show.
Returns
-------
None.
"""
fig, ax = plt.subplots(nrows=1, ncols=1, figsize=(10, 10))
ax.imshow(image) # , cmap='gray'
ax.axis("off")
plt.show()
return fig, ax
[docs]
def compare_images(title1, im1, title2, im2, suptitle=None, figsize=(10, 5), save_path=None) -> None:
"""
Parameters
----------
im1 : np.ndarray
image 1.
title1 : str
title for image 1.
im2 : np.ndarray
image 2.
title2 : str
title for image 2.
figsize : TYPE, optional
size of figure. The default is (10, 5).
Returns
-------
None.
"""
fig, ax = plt.subplots(1, 2, figsize=figsize)
if suptitle:
fig.suptitle(suptitle)
ax[0].title.set_text(title1)
ax[0].imshow(im1)
ax[0].set_axis_off()
ax[1].title.set_text(title2)
ax[1].imshow(im2)
ax[1].set_axis_off()
if save_path:
plt.savefig(save_path, bbox_inches='tight')
plt.show()
[docs]
def compare_orig_mask_gt_pred(
im: np.ndarray, mask_gt: np.ndarray, mask_pred: np.ndarray, title: str = ""
) -> None:
"""
Simple function for comparing oringal image and ground truth image
Parameters
----------
im : np.ndarray
original image.
mask_gt : np.ndarray
ground truth image.
mask_pred : np.ndarray
predicted mask.
title : str, optional
title of the plot, usually the origina filename. The default is "".
Returns
-------
None
function only just plots data.
.. image:: ./resources/visualization-compare_orig_mask_gt_pred.png
:width: 400
:alt: Grid of images showing ground truth and predicted mask, their exclusive OR results
"""
alpha = 0.5
im_overlay = label2rgb(
mask_pred, normalize(im), bg_label=0, alpha=alpha, image_alpha=1, kind="overlay"
)
fig, ax = plt.subplots(2, 3, figsize=(10, 7))
plt.suptitle(title)
ax[0, 0].title.set_text(f"original")
ax[0, 0].set_axis_off()
ax[0, 0].imshow(im)
# overlayed
dice_coeff = dice(mask_pred, mask_gt)
ax[0, 1].title.set_text(f"overlayed mask_pred")
ax[0, 1].set_axis_off()
ax[0, 1].imshow(im_overlay)
# mask gt
ax[1, 0].title.set_text(f"mask_gt")
ax[1, 0].set_axis_off()
ax[1, 0].imshow(mask_gt)
# mask pred
ax[1, 1].title.set_text(f"mask_pred \n dice: {dice_coeff:.4f}")
ax[1, 1].set_axis_off()
ax[1, 1].imshow(mask_pred)
## XOR
mask_xor = np.logical_xor(mask_gt, mask_pred)
error_total = total_error(mask_gt, mask_pred)
ax[0, 2].title.set_text(f"mask_xor\n total error: {(error_total*100):.3f}")
ax[0, 2].set_axis_off()
ax[0, 2].imshow(mask_xor)
ax[1, 2].set_axis_off()
plt.show()
if __name__ == "__main__":
import numpy as np
im = np.random.rand(40, 40)
compare_orig_mask_gt_pred(im, im, im)
# print("TODO add test code")
# im_orig = np.random.rand(512,512)
# im_gt = np.round(im_orig)
# compare_orig_mask_gt_pred(im_orig, im_gt, im_orig,"comparing originals")