small query
skrish13 opened this issue · comments
Hi,
Thanks for providing the code and making it so user-friendly! I had a small query regarding the name of the pretrained weight file shared. e121_NN_0.480_NNT_0.655_NNDR_0.813_MeanMS_0.649.pth.tar Am I right in assuming they're the accuracies using those evaluation metrics? If so, this is on which data exactly? (HPatches? Validation? Test? V/I?) Also what is MeanMS?
Also, could you share the split of train (&val) and test scenes on HPatches?
@skrish13
of course, but recently I am busy with ddl, so can you wait a few more weeks?
The file name e121_NN_0.480_NNT_0.655_NNDR_0.813_MeanMS_0.649.pth.tar comes from the code in evaluate() function:
https://github.com/Xylon-Sean/rfnet/blob/f6b54087f27caf84103806393c166c4d78d14613/train.py#L408
You could see, we call this evaluate() function on Line 423 and send a parameter called val_data
https://github.com/Xylon-Sean/rfnet/blob/f6b54087f27caf84103806393c166c4d78d14613/train.py#L423
The val_data defined on Line 214.
https://github.com/Xylon-Sean/rfnet/blob/f6b54087f27caf84103806393c166c4d78d14613/train.py#L214
Therefore, the first question is
If so, this is on which data exactly?
The answer of this question is
The evaluation data comes from line 236 to 265 in
hpatch_view.csv
The second question is
what is MeanMS?
The answer is
meanms = (PreNN + PreNNT + PreNNDR) / 3
on
https://github.com/Xylon-Sean/rfnet/blob/f6b54087f27caf84103806393c166c4d78d14613/train.py#L406
This is great! Thanks for the detailed response :)