Shape structure dataset

In [1]:

import numpy as np
from matplotlib import pyplot as plt
from curvey.shape_structure_dataset import ShapeStructureDataset

import numpy as np from matplotlib import pyplot as plt from curvey.shape_structure_dataset import ShapeStructureDataset

The 2D Shape Structure Dataset makes for a useful source of curves for testing shape interpolation.

Download the shapes zip file here. It does not need to be unzipped.

In [2]:

dataset = ShapeStructureDataset("~/Downloads/ShapesJSON.zip")

dataset = ShapeStructureDataset("~/Downloads/ShapesJSON.zip")

All classes in the dataset:

In [3]:

print(", ".join(dataset.class_names))

print(", ".join(dataset.class_names))

apple, bat, beetle, bell, bird, bone, bottle, brick, butterfly, camel, car, carriage, cattle, cellular_phone, chicken, children, chopper, classic, comma, crown, cup, deer, device, dino, dog, elephant, face, fish, flatfish, fly, fork, fountain, frog, glas, guitar, hammer, hat, hcircle, heart, horse, horseshoe, image, jar, key, lizzard, lmfish, misk, octopus, pencil, personal_car, pocket, rat, ray, sea_snake, shoe, spoon, spring, stef, teddy, tree, turtle, watch

For each class, there are (usually) multiple exemplars:

In [4]:

print(", ".join(dataset.names_by_class["apple"]))

print(", ".join(dataset.names_by_class["apple"]))

apple-1, apple-10, apple-11, apple-12, apple-13, apple-14, apple-15, apple-16, apple-17, apple-18, apple-19, apple-2, apple-20, apple-3, apple-4, apple-5, apple-6, apple-7, apple-8, apple-9

Load a curve either by its full name or its class and index:

In [5]:

dataset.load_curve("apple", 6)

dataset.load_curve("apple", 6)

Out[5]:

Curve(n=103)

Plot the first exemplar in each class:

In [6]:

n = len(dataset.class_names)
n_cols = 10
n_rows = int(np.ceil(n / n_cols))
fig, ax = plt.subplots(figsize=(12, 1.2 * n_rows))

for i, class_name in enumerate(dataset.class_names):
    c = dataset.load_curve(class_name, 0).translate("center")
    max_r = np.linalg.norm(c.points, axis=1).max()
    c = c.scale(0.35 / max_r)
    y = i // n_cols  # This is correct
    x = i - y * n_cols

    c.translate([x, -y]).plot(color="black")
    plt.text(x, 0.4 - y, class_name, horizontalalignment="center")

ax.axis("equal")
ax.axis("off")
ax.set_ylim((-(n_rows - 0.5), 0.75));

n = len(dataset.class_names) n_cols = 10 n_rows = int(np.ceil(n / n_cols)) fig, ax = plt.subplots(figsize=(12, 1.2 * n_rows)) for i, class_name in enumerate(dataset.class_names): c = dataset.load_curve(class_name, 0).translate("center") max_r = np.linalg.norm(c.points, axis=1).max() c = c.scale(0.35 / max_r) y = i // n_cols # This is correct x = i - y * n_cols c.translate([x, -y]).plot(color="black") plt.text(x, 0.4 - y, class_name, horizontalalignment="center") ax.axis("equal") ax.axis("off") ax.set_ylim((-(n_rows - 0.5), 0.75));