This week, I have created a facial recognition model using Histogram of Oriented Gradients (HOG) method.

​

You may wonder what is HOG... A brief explanation

​

It is a simple and powerful feature descriptor. It is not only used for face detection but also it is widely used for object detection like cars, pets, and fruits. HOG is robust for object detection because object shape is characterized using the local intensity gradient distribution and edge direction.

​

This is the architecture of how the HOG method will process 

from the raspberry pi camera to recognizing faces. 

​

It will start by getting the image from the camera, next

it will process the image to compute as 128-d vector face 

embedding via deep metric network. It will then compare to

the database to find a matching vector and recognize the face

of the user.

​

My idea will be implementing this into the raspberry pi so that there will not be a need for a server to be involved for ease of use. I am not confident with the confidence level of the raspberry pi as it is not meant for machine learning. I will be trying to improve the facial recognition model in the future as the prediction accuracy is at 0.36. Not reliable so I will need to cramp a lot of things next week :D

​

​

​

​