Face Detection by Avinoam David

About the Project

• This is a Real-Time Face-Detector Project based on user inputs. It features an Easy and Interactive Guide to taking self-made photos and feeding them to the detector training session.
• This project was made by me after taking several courses regarding both practical and theorem behind Computer Vision fields such as (to name a few) Convolutional Neural Networks, Image Proccessing Tools, Object Detection Tools.
• The project integrates both Classification and Regression (Prediction) models useage. I made the project to test and display my ability to put my learnings into practice.
• Images handling done with OpenCV
• Labeling done with "labelme"
• model handled with TensorFlow

Some Runtime Examples:

Bad Lighting                                         While Moving                Various Positions

Walkthrough:

The project implements a complete process of creating a working real-time object detector, demonstrated using face as the key object (although capable with any object):

1. Prepping Data:

• taking photos
• labeling with "labelme" software
• partitioning to train, val, and test
• augmenting to enhance dataset

2. Building A network:

• Using VGG16 netwrok for Classification and Integrating a self written Loss Function for Regression.
• The integration is made to be able to classify a face and predict the indicator box (around faces) co-ordinates

3. Experimenting with variables to improve the model (epochs, types of augmentations, adjustments of the loss function, etc)
4. Back-Testing and Visualising the results

Train and Use it yourself!

• Simply install all dependencies specificed in FaceDetection.ipynb (Set an environment if necessary)
• Run every cell in the correct order. When meeting a cell with a title !Practipicate! - follow the instructions. they are ment so that the model will learn based on your very self!