FrechetLib

A library for computing Frechet distances between curves, and other associated quantities. This is a Python version of the FrechetDist julia package.

Development Status

This library is in an alpha state. Most functions have some degree of testing, but not extensively.

Examples

import time

from frechetlib.continuous_frechet import frechet_c_approx
from frechetlib.data import FrechetDownloader

downloader = FrechetDownloader()
big_curve_1 = downloader.get_curve("11/poly_a.txt")
big_curve_2 = downloader.get_curve("11/poly_b.txt")

print(big_curve_1.shape)
print(big_curve_2.shape)
# NOTE Run twice to factor out the compilation time from the timing.
frechet_c_approx(big_curve_1, big_curve_2, 1.01)
print("Starting")

start = time.perf_counter()
res, morphing = frechet_c_approx(big_curve_1, big_curve_2, 1.01)
end = time.perf_counter()

print("Approximated Frechet distance: ", morphing.dist)
print("Time taken (s): ", end - start)

Benchmarks

To run the standard benchmarks, you first need to have poetry installed. After cloning the repo, run the command

poetry install --with dev

That will install the package from source in a virtual environment with the dev dependencies. To start the benchmarks, run

poetry run python paper_bench/run_experiments.py

This will then execute the benchmarks and put the results in benchmark_results.csv.

API

The package API is currently minimal and undocumented for now. See the test cases for usage examples. This will be filled out more later.