This project serves a deep learning model scoring selfie images between 1 to 10 based on image and face attributes. You can learn the technical details of this project from this blog post. Use [resnet.torch] (https://github.com/erogol/resnet.torch), if you plan to follow all the training pipeline described on the post.

Given image is processed as follows;

1. Detect face.
2. Find landmarks
3. Rotate image to align face.
4. Fill gaps with constant pixel value.
5. Send into scoring model.

For an example use check notebook ExampleUse.ipynb

• Contact me from erengolge at gmail.com

Main requirement is Torch computing framework.

dlib face model - place under utils/
beauty model - GPU model (use utils/convert2cpu.lua for setting it for CPU) place under trained/
optimstate - if you like to fine-tune the model.

dlib sudo pip install dlib - face and landmark detection)
lutorpy sudo pip install lutorpy - using torch model on python
skimage sudo pip install skimage - image processing
cv2 sudo pip install cv2 - OpenCV python module

• Face alignment code in utils/img_processing.py.
• A template for porting Torch models to python in utils/Classifier.py.
• The model itself

Attention of the trained model.

Sorting A. Lima images from Google Search.