Also save predictions in separate csv/geojson with 1 row per prediction

[rwood-97]

For viewing results, would be good to save the predictions to a csv/geojson file with 1 row per prediction instead of aggregated at the article level.

Code for doing this using current csv:

import pandas as pd
from ast import literal_eval
import geopandas as gpd
from shapely import Point

preds = pd.read_csv("./UK_France_test_sample_pretrained-perfectmatch-reldisamb-predictpub-combined-nomicro.csv")
preds["predictions"] = preds["predictions"].apply(literal_eval) # apply literal_eval to ensure predictions column is interpreted as a list

predictions = preds["predictions"].tolist() # get predictions column
predictions = [
    pred for prediction in predictions for pred in prediction
] # flatten list

df = pd.DataFrame(predictions)
df["predicted_coordinates"] = df["predicted_coordinates"].apply(Point) # convert coords to shapely point
gdf = gpd.GeoDataFrame(df, geometry="predicted_coordinates") # create geodataframe

gdf.to_file("UK_France_test_sample_pretrained-perfectmatch-reldisamb-predictpub-combined-nomicro.geojson", driver="GeoJSON") # save to geojson

If you already have the predictions list, start from flattening.