yinguobing / facial-landmark-detection-hrnet

A TensorFlow implementation of HRNet for facial landmark detection.
GNU General Public License v3.0
158 stars 40 forks source link

need your help #2

Open Manishsinghrajput98 opened 4 years ago

Manishsinghrajput98 commented 4 years ago

Hello thanks for your code. i want to train my model on nails images. how can i do this. i was convert the dataset json to .pts format. like this version: 1 n_points :4 { 328.15 136.54 332.96 174.04 370.46 175.0 375.27 135.58 } now what can i do for next steps . can i pass the direct .pts files on train.py? can i convert to pts into other format? please help !

yinguobing commented 4 years ago

Recommended steps if you are trying to transfer this project into another usage:

  1. Do they have similar problems to solve? If you want to do classifications like separating cats from dogs, this is not the project for you.
  2. Try to understand what the repo is doing. There are multiple modules in this project but depending on your purpose not all of them are necessary.
  3. Customize the code to suit your need. Define the model's inputs and outputs. Generate valid training and testing dataset. It should not be hard if you have accomplished the two steps above.

Good luck!

Manishsinghrajput98 commented 4 years ago

thanks for reply. but i only i want to how can i train my own model. i have a halen like dataset (images corresponding pts files). only i want to know . how to pass helen dataset in into train.py ? can i need to convert halen dataset into other format?

yinguobing commented 4 years ago

Then dataset.py is a good start point.

Manishsinghrajput98 commented 4 years ago

okay after dataset.py then run train.py ? right ?

yinguobing commented 4 years ago

Before training, make sure the network's inputs and outputs shape are compatible with the new dataset.

Manishsinghrajput98 commented 4 years ago

the repository based on wflw dataset. but i have a helen like dataset. so how can i convert the dataset according to train.py thanks