search

oufuzhao / EQBM

Code for Troubleshooting Ethnic Quality Bias with Curriculum Domain Adaptation for Face Image Quality Assessment
https://openaccess.thecvf.com/content/ICCV2023/papers/Ou_Troubleshooting_Ethnic_Quality_Bias_with_Curriculum_Domain_Adaptation_for_Face_ICCV_2023_paper.pdf
6 stars 2 forks source link

Dataset format #1

Open phonPtan opened 4 hours ago

phonPtan commented 4 hours ago

How can i train in my dataset? You can define the source_list, target_list, target_label, pretrain_model parameters in config_train.py

oufuzhao commented 3 hours ago

Hi,

Thank you for your interest in our work.

The "source_list" is the path of the txt file to indicate your source dataset, the format is like this: /XXXXX-Your-Source-Dataset-PathXXXXX/XXXXX1.jpg {$quality labels} /XXXXX-Your-Source-Dataset-PathXXXXX/XXXXX2.jpg {$quality labels} /XXXXX-Your-Source-Dataset-PathXXXXX/XXXXX3.jpg {$quality labels} ...

The "target_list" is the path of the txt file to indicate your target dataset, the format is like this: /XXXXX-Your-Target-Dataset-PathXXXXX/{$target_label}/XXXXX1.jpg /XXXXX-Your-Target-Dataset-PathXXXXX/{$target_label}/XXXXX2.jpg /XXXXX-Your-Target-Dataset-PathXXXXX/{$target_label}/XXXXX3.jpg ...

Please note that the "target_label" need to be included in the data path to filter out data with a specific "target label".

The "pretrain_model" is the path of the pre-trained face recognition model with the ".pth" pytorch format.