MethaneUnion

This dataset product is organized as:

datasets/
  temporal_split/
    original_scale/{train,test}.csv
    120m_GSD/{train,test}.csv
    360m_GSD/{train,test}.csv
    480m_GSD/{train,test}.csv
    960m_GSD/{train,test}.csv
  geo_split/
    original_scale/{train,test}.csv
    120m_GSD/{train,test}.csv
    360m_GSD/{train,test}.csv
    480m_GSD/{train,test}.csv
    960m_GSD/{train,test}.csv

Each CSV contains exactly these columns: id, label, latitude, longitude, available_sensor, S2_t0_path, S2_pre_path, S2_pre_pre_path, S2_plume_label_path, L89_t0_path, L89_pre_path, L89_pre_pre_path, L89_plume_label_path, EMIT_t0_path, EMIT_pre_path, EMIT_pre_pre_path, EMIT_plume_label_path, S5p_temporal_path.

Path columns point to files inside the uploaded dataset_part_XXX.tar.gz archives. S5P temporal files are .npz; plume labels are named with sensor-specific *_plume.tif filenames.