Delve into facial recognition with Python libraries MediaPipe, Dlib, and DeepFace. Explore face detection, analytics, transformation effects, and full makeovers, gaining crucial biometric software skills.

updated_FA.tar.gz

setup

Face Detection

Face Landmarking Using Dlib

Face Landmarking Using MediaPipe

Delaunay Triangulation of Facial Landmarks

Face Alignment

Age Prediction

Gender Prediction

Emotions Prediction

Race Prediction

Recognizing Eyes Color

Calculating Eyes Open Ratio

Detecting Eyeglasses

Recognizing Lips Color

Changing Lips Color

Detecting Skin Color

Swapping Faces

Swapping Eyes

Drawing Face Filter

Blurring Face

Cartooning Face

Face Encoding

Face Matching - Manhattan

Face Matching - Euclidean

Try Virtual Glasses

Change Background

Expand Face Features

Remove Background

Predict Beauty

Detect Face Shape

Condense Eyebrows

Drawing Eyeliners

Applying Blush

Detect Hair Color

Change Hair Color

Aging Faces

Auto-colorize Facial Photos

Calculate Face Golden Ratios

Add AR effects To Facial Photos

Face analysis technology is a rapidly growing biometric software discipline with wide-ranging applications in surveillance, forensics, game design, and social media. As with other machine learning domains, Python has several libraries for computer vision, image analysis, and pattern recognition that make it ideal for facial analysis.

This course is a hands-on introduction to facial recognition with three unique libraries—MediaPipe, Dlib, and DeepFace. You’ll start with face detection, landmarking, and face alignment before exploring common analytics like age, gender, and emotional prediction based on facial expressions. Next, you’ll identify common facial features before transforming them by adding blurring, sketching, and cartoon effects or swapping color palettes. You’ll finish by performing full makeover functions, manipulating cheeks, lips, eyes, and brows.

By the end of this course, you’ll have a strong foundation in popular facial recognition and manipulation libraries in Python.

Analyzing and Transforming Faces in Python

# Introduction ##

Our hair is a salient element of our overall appearance and plays a key role in facial analysis. Accurate detection and segmentation of the hair region under unconstrained conditions remains challenging for the following reasons:

* Poor lighting conditions
* Crowded background
* Hairstyle and color variations


The segmentation of the hair region provides cues during the face recognition process and is valuable to many domains like:
* Augmented reality applications, such as hair modeling.
* Biometric recognition applications, such as detecting the presence of a human when their back is turned to the camera.

# Objective ##

This lesson will explore how to exploit a pretrained ONNX model for hair segmentation. We’ll develop a Python utility that detects the color of the segmented hair region.


# Dependencies ##

We’ll be using the following external Python libraries.


align-center

# Let’s code this utility! ##

Before exploring the core functions of this utility, let’s get introduced to ONNX.  **ONNX** is an open standard format for neural network model interoperability. It allows the transfer of heavy, trained neural networks from one training tool to another (for example, between PyTorch and TensorFlow).  It also deploys models to the cloud using the **ONNX Runtime**.


Using ONNX for predictions involves the following steps:

1. Download the relevant ONNX model.
2. Understand its inputs and outputs.
3. Preprocess the data (for example, an image) based on the required input format.
4. Perform inference using ONNX Runtime.
5. Visualize predictions.


Now, we’ll go through the main functions of this utility.

## The `initialize_hair_segmentation_model` function ###

This function initializes the ONNX model and starts an inference session. The function allows us to perform the following steps:

* **Line 5** instantiates an inference session. This session is used to load and run the ONNX model as well as to specify environment and configuration options.
* **Lines 10–16** query the ONNX model metadata and get the definition of its input metadata.
* **Lines 19–25** query the ONNX model metadata and get the definition of its output metadata.


The complete code of the function `initialize_hair_segmentation_model` is shown below:

Let’s learn how to detect hair color in a digital image.


Library	Version
onnxruntime	1.9.0
MediaPipe	0.8.9
opencv-python	4.4.0.46
scikit-learn	1.0.1
NumPy	1.19.4
webcolors	1.11.1
filetype	1.0.7

Introduction

Core Functions

Predictive Analytics

Manipulation Functions

Virtual Makeover Functions

Face Recognition

Conclusion

Appendices

Detect Hair Color

Introduction

Objective

Dependencies

Let’s code this utility!