Face Classification based on Python

Background

This is the summer semester homework for THU EE freshman in 2021.8, the purpose of which, interpreted by the consoulors, is to get familiar with the basic grammar of Python. Therefore, the accuracy of the gender classification is not what to be concerned with and all methods and algorithms, no matter how basic or advanced they might be, are encouraged.

Intension

The very primary intension of creating this repository is to check out whether I've learnt the basic using skills of git and Github rather than others.
Meanwhile, the repository might be a help for the junior if Python lectures is to be remained during the coming summer semester.

Branches

There are two branches separated base on different environments.

master

Files
- cnn.py (Runnable)trains the Convolutional Neural Networks and print the accuracy of the face classification;
- Common contains the basic classes used in cnn.py, including dataloader, callback, etc.
- Dataset contains the image data used for training. Image contains the original forms and Label the correct gender. To improve accuracy, they are pre-treated and are saved in folder Image-haired and Image-haired_colored.
How to use

Clone the branch to your PC and open the entire folder with VS code, and the py file cnn.py can be ran directly.
Environment

Tensorflow 2.0
Algorithm

CNN

base

Files
- KNN.py (Runnable)print the accuracy of the face classification based on KNN;
- KNN-SKLearn.py (Runnable)print the accuracy of the face classification based on SKLearn KNN which is faster;
- LogisticRegression-SKLearn.py (Runnable)print the accuracy of the face classification based on LR, which is more accurate than KNN;
- Common contains the basic classes used in cnn.py, including dataloader, classifier, etc.
- Dataset contains the image data used for training. Image contains the original forms and Label the correct gender. To improve accuracy, they are pre-treated and are saved in folder Image-haired and Image-haired_colored.
How to use

Clone the branch to your PC and open the entire folder with VS code, and the py files KNN.py, KNN-SKLearn.py and LogisticRegression-SKLearn.py can be ran directly.
Environment

Python 3
Algorithm
- KNN;
- Logistic Regression.

Results

Results are based entirely on the accuracy of classification:

95% for CNN
80% for KNN
85~90% for LR.

GiddaZhang / Face-Classification