mars-cloud-data-grouping

The dataset uses a file_rad identifier to link citizen science observations to expert labels. Since evaluations must happen per image:

1. Load CSVs using pandas.
2. Pre-group the DataFrames using .groupby('file_rad') and store them in a dictionary for fast lookup.
3. Ensure that the evaluation loop iterates over all file_rad values present in the expert dataset to account for missed detections (FNs).

import pandas as pd

citsci = pd.read_csv('citsci_train.csv')
expert = pd.read_csv('expert_train.csv')

# Grouping for efficient access
cit_groups = {name: group[['x', 'y']].values for name, group in citsci.groupby('file_rad')}
exp_groups = {name: group[['x', 'y']].values for name, group in expert.groupby('file_rad')}

all_image_ids = expert['file_rad'].unique()