Results after inferece
Hi!
I managed to run the training and inference for track/shower classification, but when analysing the outcome I saw some weird results. The scores for the neutrinos seem to be quite weird, with the track_score for all of them being nearly zero. These are the scores for numuCC
On the hand, the scores for the muons seem to be correct
I do not think it is a training issue, as for the training the accuracy did seem to be correct
Epoch | LR | train_loss | val_loss | train_acc | val_acc
------------+-------------+-------------+-------------+-------------+------------
1 | 0.025 | 0.318956 | 0.563796 | 0.84375 | 0.768423
2 | 0.0245 | 0.219524 | 0.455009 | 0.9375 | 0.832758
3 | 0.02401 | 0.304384 | 0.430548 | 0.84375 | 0.829973
4 | 0.0235298 | 0.471608 | 0.44156 | 0.875 | 0.85177
5 | 0.0230592 | 0.568041 | 0.406209 | 0.8125 | 0.849387
6 | 0.022598 | 0.451703 | 0.424125 | 0.8125 | 0.855117
Do you have any idea on what could be causing this? Hope there is enough information!
Thanks in advance!
PS: I found out there is the very same issue with the output files from orcanet predict.
Edited by asanchezcastillo