CPM

Pattern Makeup - Training Guideline

This is a training guideline for Pattern Branch (P), one out of two branches in Color-Pattern Makeup Transfer (CPM)

---

1. Requirements: Please refer to Getting Started/ Requirements, the main components are:
   • segmentation_models.pytorch
   • torch >=1.6
2. 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
   

3. Training/ Evaluation: python train.py --datapath /pathtodata. Test result should be found in Tensorboard.
4. (Optional):
   • Open Tensorboard: tensorboard --logdir=runs
   • Change backbones, pretrained_weights: Check parser.py

Snapshot of Tensorboard at epoch 1. Evaluation metrics on test set should be found (Accuracy, Recall, Precision, etc.)

Acknowledgements

This code is based on PRNet and segmentation_models.pytorch

This site is open source. Improve this page.