All training pairs from different signers have been ignored #3

New Issue

2019-09-08T17:08:48+08:00

chendongliang87 commented

2019-09-08 17:08:48 +08:00

(Migrated from github.com)

I noticed in your dataset_paris.py line 56 has ignored all training pair which label=2. So it means that you didn't train and test your model with signatures from different signers. Any specific reason for this?

        if (only_label is not None and label != only_label) or label == 2:
            continue

I noticed in your `dataset_paris.py` line 56 has ignored all training pair which `label=2`. So it means that you didn't train and test your model with signatures from different signers. Any specific reason for this? ``` if (only_label is not None and label != only_label) or label == 2: continue ```

dlutkaka commented

2019-09-09 11:56:59 +08:00

(Migrated from github.com)

it's for further research，just ignore it.
lable2: means 'random_forgery'

it's for further research，just ignore it. lable2: means 'random_forgery'

chendongliang87 commented

2019-09-10 10:00:57 +08:00

(Migrated from github.com)

okay understood. but if not training with random_forgery, will the model converge well? Will it improve AUC if add random_forgery into the training set and mark as 0?

okay understood. but if not training with `random_forgery`, will the model converge well? Will it improve AUC if add `random_forgery` into the training set and mark as `0`?

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: dearsky/DeepHSV#3