Facial-Recognition-App

AI-based Face Recognition Web Application with Flask & Deployment: A facial recognition deep dive into greyscale, image processing with OpenCV, Eigen images & theory. Utilizing Python, classification with SVMs, Flask (Jinja Template, HTML, CSS, HTTP Methods), pipeline model, Heroku & more.

The model development process involves creating a streamlined pipeline for data preprocessing, analysis, model training, and parameter tuning.

Once developed, the face recognition model is integrated into a Flask application and then deployed to Heroku. The entire project is structured to ensure an end-to-end understanding of developing and deploying a machine learning-based web application, starting from scratch.

A snapshot of the process

Downloading dependencies:

EDA

Mapping out the image sizes

Gender distribution

Exploring Grayscale for Facial Recognition

Creating a grayscale color map

Getting OpenCV, one of the most facial recognition modules

Testing face detection

We will now be extracting female and male faces and putting them into their respective categories

Cropping images & grayscale

Using PCA & Eigen Images

Using GridSearchCV for looping through params for the ML model

Checking ROC Curve for Model

Creating dataframe to represent the ML

Unleashing Facial Recognition on Video

Known issues

• Some issues that can be addressed in future projects:
  • Bright spots are considered faces sometimes
  • Some side faces not being picked up as a face
  • Not every frame is picked up as a face (but this could depend on frame rate and tuning)
  • Some faces with strange masks on (i.e, Captain America mask) is not being detected as a possible face
  • A very-much covered face is undetectable, is it possible to create face detection with question marks for a human to review

Not all side-faces go un-detected

Example of masked face going un-detected

Creating the Flask App base template:

Gender classification code for the Flask App:

Flask App

Beginning

Homepage code

Home page, on the Flask app

After some basic styling, Flask app look & feel:

Testing the front-end gender classification model

Choosing file:

Execution & Result:

Example of more results for females:

Gender Classification for males: