| name | mars-cloud-data-grouping |
| description | Organizing annotation data by image identifiers to prepare for image-by-image clustering analysis. |
The dataset uses a file_rad identifier to link citizen science observations to expert labels. Since evaluations must happen per image:
- Load CSVs using
pandas.
- Pre-group the DataFrames using
.groupby('file_rad') and store them in a dictionary for fast lookup.
- 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')
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()