This is a training guideline for Pattern Branch (P), one out of two branches in Color-Pattern Makeup Transfer (CPM) |
- Requirements: Please refer to Getting Started/ Requirements, the main components are:
- segmentation_models.pytorch
- torch >=1.6
- Data Preparation: Please download CPM-Synt-1 in here, or via wget:
wget https://public.vinai.io/CPM-datasets/CPM-Synt-1.zip unzip CPM-Synt-1.zip
- Training/ Evaluation:
python train.py --datapath /pathtodata
. Test result should be found in Tensorboard. - (Optional):
- Open Tensorboard:
tensorboard --logdir=runs
- Change backbones, pretrained_weights: Check
parser.py
- Open Tensorboard:
Snapshot of Tensorboard at epoch 1. Evaluation metrics on test set should be found (Accuracy, Recall, Precision, etc.) |
This code is based on PRNet and segmentation_models.pytorch