- To look for elves, inside a directory with Mini-EUSO root files call
find_elves_bigsample
it will produce tle_occnn_results_bigsample0.txt file with filename, frame and distance to the ELVESs hyperspace centre. The smaller the distance (usually <0.1), the bigger the chance it is an ELVES. But sometimes ELVESs have also pretty high distances...
- Another model to look for ELVESs:
find_elves_smallsample
this model in principle gave worse result than the one above, but it was also tested to generalise quite well, while the one above wasn't.
- Sun filtering of the found ELVES
A lot of found events are not ELVESs but sun influenced data. If you are operating on S1 data, one can create filtered lists of ELVESs with
filter_sun_tles_occnn.py
- Browing through the elves that were found with etos: modify the file below to suit your directories:
browse_tles_tmp.py
- Training is far more complicated
a. You need to generate samples from Mini-EUSO data and augument them like in generate_samples.py b. You need to run training (after modyfying absolute paths and the initial model) with train_elves_occnn.py