Strava Art

Ultimate goal is to design routes as the contour of an image.

How to convert an image to a Route

Here is the recipe:

1. Extract the largest contour (list of coordinates) from an image
2. Project that contour on a map.
3. Get direction between successives points.

Have a look at the Starter Notebook for a first overview.

from stravart.polygone import Polygon
from stravart.search.operations import Projection
from stravart.contours.extraction import ContourExtractor

contour_extractor = ContourExtractor("path/to/img")
contour = contour_extractor.get_best_contour()
contour.close()

map_center =  (48.8675, 2.3638)  
radius = 0.03

poly =  Polygon.from_list(coordinates_list=contour, system="cartesian")
projection = Projection(center=map_center, radius=radius, map_type="GPS")
gps_poly = projection.apply(poly)

final_contour, path_mapping = gps_poly.fill_paths_between_points()
plot_route(map_center=map_center, route=final_contour)

For natural image, check out the Contour Extraction.ipynb notebook.

Find a better route

We can try to look for the best possible route by:

1. Applying multiple operations to the polygon (dilatation, rotation)
2. Measure each feasible route quality - as the total difference area between the image contour and the suggested polygon.
3. Select the best one according to the previous criterion.

Check out the Optimization notebook

import optuna
from stravart.search.optimization import generate_grid, objective

# Grid over Paris
lat_start, lat_end = 48.8156, 48.9022
lon_start, lon_end = 2.2241, 2.4699
city_grid = generate_grid(lat_start, lat_end, lon_start, lon_end, 10, 10)

study = optuna.create_study(direction='minimize')
study.optimize(lambda trial: objective(trial, poly=poly, city_grid=city_grid), n_trials=30)

best_trial = study.best_trial
best_final_contour = best_trial.user_attrs['final_contour']
map_center = city_grid[study.best_params['map_center_idx']]

Example

For this image of a dog

Running the code above (contour extraction + optimization) results in this route which is around 30kms long:

Installation

To install StravArt, follow these steps:

1. Clone the repository to your local machine

git clone https://github.com/yourusername/StravArt.git
cd StravArt

2. Create a virtual env

python -m venv venv
source venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

4. Set up API access tokens By default, we use Google Maps API. You need to get an access token first and set it as environment variable.

export GMAPS_KEY=youraccesstoken

or directly in Python

import os
os.environ['GMAPS_KEY']='youraccesstoken'

TODO

• Check out SAM-like approach directly in a webapp ? (check out this one in streamlit or this blog)