Eye Center Identification

Image recognition has become a popular and challenging topic in the field of machine learning.This project is a symplified version of Kaggle's competition--Facial Keypoints Detection.

This project applies machine learning classification techniques to make predictions of the eye center locations on facial images. A GUI application has been made to show the predicted results.

Data

1. Original data source: https://www.kaggle.com/c/facial-keypoints-detection/data training.zip
2. Data used in this project: the first 200 samples from the original data set

Language

The programs in this project are written in Python3.5.

Programs and Usage

1. my_func.py

   • description: functions for data processing and results visualization
   • usage: called by BuildModel_entire.py, best_model.py

2. image_preprocess.py

   • description: functions for image transformation
   • usage: called by BuildModel_entire.py, best_model.py

3. eye_identifier.py

   • description: EyeCenterIdentifier class and GridSearch class
   • usage: called by BuildModel_entire.py and best_model.py

4. BuildModel_entire.py

   • description: processes the data, builds EyeCenterIdentifier and does GridSearch
   • called using command line:
     ipython3 BuildModel_entire.py [transformation] [clf] transformation can take values: none, histeq, derivative clf can take values: LogisticRegression, RandomForestClassifier, SVC, SGD e.g ipython3 BuildModel_entire.py none LogisticRegression

5. best_model.py

   • description: builds the "optimal model" suggested by the GridSearch

6. build_model_gui.py

   • description: a model class including methods for data processing, model building and results predicting
   • usage: called by the eye_center_gui.py program

7. eye_center_gui.py

   • description: a GUI which processes the data, builds the model and enables the users to select facial images for the predictions
   • called using command line:
     ipython3 eye_center_gui.py

Example

Here is an example of the predictions of eye locations:

The red dots represent the predicted eye center locations, the green dots represent the true eye center locations and the blue dots represent the results from the benchmark model.

GUI

To see the demo of the GUI application, please click here.