Giters

CAMMA-public / HPE-AdaptOR

Code for Unsupervised domain adaptation for clinician pose estimation and instance segmentation in the OR

Home Page:https://arxiv.org/abs/2108.11801

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Strong and Weak Transforms

yaldashbz opened this issue · comments

commented

Hi,

I couldn't find which transforms you used as strong and weak transforms for source and target datasets.
Could you please specify them?

Thanks.

commented

Hi, here is the data mapper class that applies the weak and strong transformations:

HPE-AdaptOR/adaptor/data/dataset_mapper_weak_strong.py

Line 185 in 08a3b8b

if self.apply_rand_augment and self.apply_low_res_resize:

settings for the transformations:

HPE-AdaptOR/configs/base/Base_kmrcnn_R_50_FPN_3x_gn_amp_4gpus.yaml

Line 62 in 08a3b8b

APPLY_RAND_AUGMENT: True