+lut¶
These files were created for SCOPE per pixel implementation within the TRUSTEE project (H2020, TRuStEE, MSCA-ITN-2016 No 721995) tutorials. Briefly, it creates a look up table (LUT) and compares two types of search within LUT - based on root mean squared error (RMSE) and gaussian process regression (GPR).
LUT input sampling within parameter borders adjusted in src/+lut/input_borders.csv
- generate_lut_input(tab, n_spectra, outdir)¶
params = generate_lut_input(tab, n_spectra, outdir)
SCOPE run in time series mode with lut_in.csv data …
Training and testing of gaussian process regression (GPR)
- train_gpr(lut_in_path, lut_out_path, var_name)¶
Preparation of data: flattening of tiff files to csv.
- image2csv(im_path, outdir)¶
Preparation of validation data (random pixels from flattened image)
- pick100(full_set_path, n_pixels)¶
SCOPE run in time series mode with validation.csv data …
Validation of methods
- validate_gpr(gpr_path, val_in_path, val_out_path)¶
- validate_rmse(lut_in_path, lut_out_path, val_in_path, val_out_path)¶
Usage of methods
- use_gpr(gpr_path, full_set_path)¶
- use_rmse(lut_in_path, lut_out_path, full_set_path, var_name)¶
Supporting functions
- change_detection(rmse_im_path, gpr_im_path)¶
- compress_geotiff(path_in, path_out)¶
- csv2image_plane(val_in, res)¶
to read r, c from column name ind_r_c
- lut_search(params, lut_params, response)¶
check validity: all lut params in params
- plot_1to1(meas, mod, what, out_path, var_name)¶
- plot_image(im, what, out_path)¶
- write_tiff(im, out_path)¶