add script to measure snic performance

pysnic/examples/timing.py

@@ -0,0 +1,44 @@
+from pkg_resources import resource_stream
+from PIL import Image
+import matplotlib.pyplot as plt
+import numpy as np
+import skimage.color
+from skimage.segmentation import mark_boundaries
+from timeit import default_timer as timer
+from itertools import chain
+from pysnic.algorithms.snic import snic, compute_grid
+from pysnic.ndim.operations_collections import nd_computations
+from pysnic.metric.snic import create_augmented_snic_distance
+import math
+
+# load image
+color_image = np.array(Image.open(resource_stream(__name__, "../data/orchid.jpg")))
+number_of_pixels = color_image.shape[0] * color_image.shape[1]
+
+# SNIC parameters
+numSegments = 100
+compactness = 10.0
+
+# compute grid
+grid = compute_grid(color_image.shape, numSegments)
+seeds = list(chain.from_iterable(grid))
+seed_len = len(seeds)
+
+# choose a distance metric
+distance_metric = create_augmented_snic_distance(color_image.shape, seed_len, compactness)
+
+
+start = timer()
+
+segmentation, distances, centroids = snic(
+    skimage.color.rgb2lab(color_image).tolist(),
+    seeds,
+    compactness, nd_computations["3"], distance_metric,
+    update_func=lambda num_pixels: print("processed %05.2f%%" % (num_pixels * 100 / number_of_pixels)))
+
+end = timer()
+print(f"superpixelation took: {math.floor((end - start)*1000)}ms")
+
+plt.figure("SNIC with %d segments" % len(centroids))
+plt.imshow(mark_boundaries(color_image, np.array(segmentation)))
+plt.show()