kraina-ai / srai Goto Github PK

Core libraries:

• Geospatial
  • GeoPandas
  • Pandas (explicit versioning)
  • Numpy (explicit versioning)
  • Shapely (explicit versioning)
  • H3 (4.0 is in alpha, 3.7.4 is current version)
• Utility
  • tqdm (alive-bar as alternative)

Test useful libraries for parallelization:

• numba?
• dask?
• celery XD
• ray
• swifter
• pandarallel

Optional dependencies:

• Visualization
  • seaborn (optional)
  • leaflet
• ML / AI - training
  • Scikit (?)
  • PyTorch
  • PyTorchGeo
• OpenStreetMap (test what will be useful)
  • OSMPythonTools
  • osmnx

Development:

• black
• isort
• flake8

Docs (select one):

• sphinx
• quattro
• MkDocs (fastapi uses this one)
• diagrams library (generating diagrams as a code)

Add useful libraries:

• PyFunctional
• more-itertools (add only when needed)

CodeFactor found an issue: Duplicate Code

It's currently on:
tests\test_voronoi_regionizer.py:51-65
tests\test_voronoi_regionizer.py:93-107
tests\test_voronoi_regionizer.py:119-133
Commit 262d118

Setup logging and rich library

Loads geoparquet file (or files)

Basic regionizer that uses the whole shape as a single region. Noop equivalent.

Divides given shape into H3 cells of selected resolution.

https://squidfunk.github.io/mkdocs-material/setup/adding-a-git-repository/#revisioning

https://github.com/srai-lab/srai/blob/0f53902c2401ee2bdcdb1bdc333e5070083cd73a/srai/regionizers/voronoi_regionizer.py#L42

Divides a region into Thiessen polygons (Voronoi cells) based on a list of seeds.

We should probably ask for some help from the designer.

Converts assigned objects into vector with counts of different types of features for each region.

Add integration with automated code review, stale pull requests, license checking and maybe more.

Create a module (maybe named simply models?) for quick and easy vectorization of data for machine learning.

Allow users to just pass a list of (lon, lat) or shapely.geometry.Point values in wgs84 CRS and return a NumPy array.
No GIS knowledge is required, just enrich your data with predefined models. May download the existing model from the server or run a predefined pipeline in the background with a progress bar and logging.

Examples to follow the API:

word2vec

from gensim.models import Word2Vec
word2vec = Word2Vec(all_words, min_count=2)

embeddings

from embeddings import GloveEmbedding, FastTextEmbedding, KazumaCharEmbedding, ConcatEmbedding

g = GloveEmbedding('common_crawl_840', d_emb=300, show_progress=True)
f = FastTextEmbedding()
k = KazumaCharEmbedding()
c = ConcatEmbedding([g, f, k])
for w in ['canada', 'vancouver', 'toronto']:
    print('embedding {}'.format(w))
    print(g.emb(w))
    print(f.emb(w))
    print(k.emb(w))
    print(c.emb(w))

Ease writing docstrings in VSCode and enforce them

Extends #23

Depends on #5 and vice versa. We have to pick documentation framework compatible with the docstring format.

Loads a .pbf file with OSM data.

Create a document with commands required to run when cloning repository (pre commit hooks, creating venv, downloading requirements and other things)

Loads geojson file (or files)

Joins loaded geo data with regions (assigns objects to regions based on intersection)

It might be useful to make gdf optional (default = None) and make it automatically regionize whole Earth in that case instead of requiring a manual creation of a bounding box.