Deeppixbis extractor
Added the extractor for DeepPixBiS, since it was not present here, but only in the following paper package: bob.paper.deep_pix_bis_pad.icb2019
I guess it's better to have it here as well !
Merge request reports
Activity
@heusch Sure, This portion can be cleaned up
https://gitlab.idiap.ch/bob/bob.learn.pytorch/blob/deeppixbis-extractor/bob/learn/pytorch/extractor/image/DeepPixBiS.py#L60I had models trained with another script at that time.Also, is there any change in behaviour when you change the return type of
score?added 1 commit
- 887d9e54 - [extractor] modified the code to load the model, to be consistent with MCDeepPixBiS
-
Hi @ageorge,
So I checked the modified version of the score format (it's an array of size 1x1 instead of a float) within the PAD pipeline and obtained score files are identical, so I guess you can merge !
Thanks !
41 42 # model 43 self.transforms = transforms 44 self.network = DeepPixBiS(pretrained=True) 45 self.scoring_method = scoring_method 46 self.available_scoring_methods=['pixel_mean','binary','combined'] 47 48 logger.debug('Scoring method is : {}'.format(self.scoring_method.upper())) 49 50 if model_file is None: 51 # do nothing (used mainly for unit testing) 52 logger.debug("No pretrained file provided") 53 pass 54 else: 55 logger.debug('Starting to load the pretrained PAD model') 56 try: -
@heusch this part is not needed anymore right?
mentioned in commit 360ebedb
mentioned in issue bob#259 (closed)