The first part of the project consist in dectecting the global
shape of your meal. Indeed, it's a very important step of the
process, because the IA is going to analyse all of the elements
within your plate, to define wether we are facing a soup,
pastries, cakes, etc. With machine learning we can train and
teach an artificial intelligence to recognize images.
Then, the second step of the process is to isolate every part of
the plate thanks to color and texture. We can do a bridge
between facial recognition and our project. Every part of the
dish is similar to an organ of your face. Except that here, we
have colours and different textures.
According to Konstantin Poklonskiy :
"The basic flow of recognition image contains several
following steps: load data, transform data according to
requirements of each model, load model and create predictor,
make a prediction, and map labels with estimation results of
the network."
For further information, it's kindly ask to visit the next
website
: Image recognition with MobileNet and ML.NET.
Then, to developp our application we will use ML5, (Click here
), it's an open source Machine Learning Library (based on
JavaScript) for the Web. It will give us our guideline to
developp our technology. Alongside ML5, we will use Tensorflow
(Click here)
. In addition, we will use MobileNet, which is a convolutional
neural network, trained with samples of images. It will help us
to construct the basic requirements of our application. But,
during a second phase, we will have to train our apps with our
own dataset of images, in order, to make it more efficient and
detect a broader range of images. To manage our work, we will
need to take an important numerous of the same picture. It will
be our raw data. Several position of the camera will be required
to detect every texture, shape or elements, which are necessary
for our apps to understand what kind of food we will face.
Our main course regarding ML in JS
For farther information about the process, it's recommmended to
read the following document: (Click here
). This is an article from MIT which explain the process and
which allow us to understand how does it works.
To sum up...
The main idea of our project is to help you to find your
Madeleine de Proust...We strongly believe that sense are
connected to experience and life. So that, you will be able
to feel again the spirit of your childhood. Thus, our
partnership with stores will help you find the best
products, at the best price.
