Overview

This repository contains a code to train a facial keypoint detection using pretrained resnet models in PyTorch.

Requirements

Once the repo cloned, please run inside the repo:

import os
os.mkdir('data')
os.mkdir('checkpoints')

Then move the training csv file to data folder.

Training

To train the model, please run:

python -m src.train 
    --batch_size [BATCH SIZE value] 
    --lr [LEARNING RATE value] 
    --wd [WEIGHT DECAY Value] 
    --train_device [choose cpu or gpu]
    --test_device [choose cpu or gpu]
    --modelname [MODEL: 'resnet18' or 'resnet50']

All other parameters can be chosen in src/training_parser.py.