How/Where to start image processing for Face Recognition using Machine Learning & Deep Learning

 
Demo: Face Recognition Using Deep Learning

There are unlimited possibilities of Face Recognition or Image Processing using M.L & Deep Learning some of applications we use in our daily life Snapchat but these all apps rely on simple face detection for recognition you need lot more processing and some experience in open source technologies.

What is required to recognize some pretty faces?

1. Ubuntu 16.04: Some experience with terminal & basic shell scripts.
2. Python 2.7/3.5: Mid-level coding & debugging skills in this open-source language for importing big Machine Learning packages into a simpler way.
3. Some research about images, processing, servers & how to use command line arguments. 

I will start from the beginning as when I was a usual person using windows & trying to execute codes for my basic projects & applications. If you are already experienced in Linux just skip to step: 3

1. 1. Dual boot your system with Ubuntu 16.04 LTS, because when you reach to last steps of compiling FaceRecognizer classes from source of opencv you will get lots of errors in earlier or later versions of Linux.
2. 2. By default you will get lot's of pre-packaged configurations with GNOME desktop flavour & I personally love this one if you are working on server side you must install XFCE4 light-weight desktops.
3. 3. Install opencv & necessary packages: I have learnt lots of things from past 2years & if you really want to learn in more practical way read all his tutorials. Link: http://www.pyimagesearch.com/2016/10/24/ubuntu-16-04-how-to-install-opencv/
4. 4. Why to build from source & include opencv_contrib module? Answer is there are lots of opencv bindings & classes which opencv officially didn't provide & the main FaceRecognizer class. So, it's strictly recomended you build from source in virtualenv else you will get error after hours of compiling & personally I have tried in lot's of ways, so stick to tutorial. If you want to read official documentation. Link: http://docs.opencv.org/2.4/modules/contrib/doc/facerec/facerec_tutorial.html#face-recognition-with-opencv
5. 5. Solving build & compile errors: At, some point after 83% compilation you may fatal error: hdf5.h: No such file or directory. Follow this issue for details: https://github.com/opencv/opencv/issues/6016. Simply, include these lines find_package(HDF5) include_directories(${HDF5_INCLUDE_DIRS}) before 3rd line in your downloaded/unzipped file opencv/modules/python/cmake.common

Where to learn or find codes of Face Recognition?

• -Face Detection Tutorials: https://www.youtube.com/watch?v=88HdqNDQsEk, Source Codes 
• -Face Recognition based on LBPH algorithm this has lot's of constraints specially if you are building for client-server application like camera/face angle, lighting conditions, frame rates etc. in general it has nearly 60-70% efficiency: https://www.youtube.com/watch?v=4W5M-YaJtIA, Source Codes 
• -Face Recognition using Deep Learning: https://github.com/ageitgey/face_recognition
• -Tutorial: https://medium.com/@ageitgey/machine-learning-is-fun-part-4-modern-face-recognition-with-deep-learning-c3cffc121d78
  

You will find lots of source code but I will recomend learn Machine Learning before compiling these codes then only you will be able to understand what is the math behind this. Learn here: https://www.coursera.org/learn/machine-learning

What after all these? What applications I can build?

• -Google Vision API for object recognition: https://www.youtube.com/watch?v=eve8DkkVdhI
• -Optical Character Recognition: https://www.pyimagesearch.com/2017/07/10/using-tesseract-ocr-python/
• -Automatic Number Plate Detection: https://www.youtube.com/watch?v=fJcl6Gw1D8k
• -Raspberry Pi Smart Robots
• -Build robots that can follow traffice rules & self-driving cars. Learn here: https://in.udacity.com/course/self-driving-car-engineer-nanodegree--nd013
  

10:03