There should be 2 folders in this directory.

|-01FaceDetection

|-02Model

01FaceDetection

This folder contains the code to detect the faces in an image, crop it and save the cropped image.

Please refer to the README.md in that folder for more information.

02Model

This folder contains the code to train the model and perform validation.

Instructions

<------------------------------ QUICKEST GUIDE ------------------------------>

1> Quick Validation

We have put some cropped and beautified images in ./02model/data/EECS442_Makeup_Go/result_test. In order to get the model's output on those beautified images, change directory to ./02Model and run the following code

python test.py -lp test

You should be able to see the output under ./02model/output/.

<------------------------------ QUICKEST GUIDE ------------------------------>

2> Makeup-Go on your own image

In order to start from the beginning, you should prepare a portrait and put it in <./01Dataset/faces> and run

python3 detect_crop.py

This will detect the face in the portrait and crop the image to the desired size. The output will be in ./01Dataset/result.

Then, you can use your own image beautification APP to perform whitening and skin-smoothing on the cropped image.

In order to perform Makeup-Go, put the beautified image in ./02model/data/EECS442_Makeup_Go/result_test, change directory to ./02Model and run the following code

python test.py -lp test

You should be able to see the output under ./02model/output/.

3> Train your own model

In order to train your own model, please change directory to ./02model and run

python train.py -lp XX<XX.yml in param/> -op XX.XX=XX<optional>

For example:

python train.py -lp train -op network.kwargs.nonlinear=ReLU,optimizer.name=Adam

We have included a small portion of training dataset so that you can run the code. If you want to train on your own dataset, here are some high level steps.

i) Face Detection and Image Cropping using the instructions in ./01FaceDetection

ii) Beautify the cropped images.

iii) Put the original cropped images and beautified cropped images in the same directory but under two folders.

iv) Copy the ./02Model/PCA_preprocess.py to that folder and run

python PCA_preprocess.py -o <original_image_directory> -b <beautified_image_directory>

And copy the yielding .t files to ./02Model.

v) Copy the original images and beautified images to ./02Model/data/EECS442_Makeup_Go/result_original and ./02Model/data/EECS442_Makeup_Go/result_beautified respectively.

vi) Run the training code

python train.py -lp XX<XX.yml in param/> -op XX.XX=XX<optional>

For example:

python train.py -lp train -op network.kwargs.nonlinear=ReLU,optimizer.name=Adam

Questions

If you have any question, send an email to [email protected] and we will respond to you asap. Enjoy.