AI into the wild

11 Nov 2018

With the project “AI into the wild” the goal to achieve was testing and bringing on their limits the solving capabilities of a CNN (Convolutional Neural Network) in a Face Recognition task.

What is a CNN?

CNN is a convolutional neural network, a program that can learn and distinguish, thanks to machine learning, a set of images and categorize them through features extrapolated from a Data-set, following the user choices.

What is AI into the wild?

The program takes advantage of two heavily used AI frameworks, Tensorflow and OpenCV. The project consists of realizing a convolutional network using Tensorflow that can extrapolate some features from the images, OpenCV to find faces from an image and at the end through transfer learning give to the convolutional network the ability to categorize the various faces.

How doew the program work?

At the beginning the program asks the user a 20 seconds video for every person to recognize. Then the neural network is trained by using the dataset (about 200 images extrapolated from the video) and gives an accuracy index. Now it’s all about throwing an image with one or more people and the algorithm will output the same image with a box about recognized faces with the name written below and the accuracy percentage.

Project Members

Simone Caldarella (Referent) simone.caldarella.it@ieee.org

Massimo Bono (Observator)