Development of lightweight AI on an embedded system

Artificial intelligence (AI) has received considerable impetus in recent years, thanks in particular to the use of graphics cards and the ever-increasing capacity of our computers. A new challenge is to be able to embed these intelligent algorithms on systems with restricted capacities.
The course focuses on three main areas:

Brief overview of different AI techniques and architectures, with particular emphasis on neural networks.
Use and creation of Docker containers with installation of the Keras framework and its dependencies.
Use and creation of neural network models in Keras.

The complete development of a lightweight application for recognizing handwritten characters captured from a camera will be considered. In order to achieve this, it will be necessary to create its own database and to determine which data will be used for training and which for testing the algorithm's effectiveness.
The implementation of the neural network training phase will be carried out on a computer using a docker container with the libraries needed to use Keras (AI framework). Once the learning model has been defined, the inference (or production) phase can be implemented. This will be coded in C language and embedded on a Raspberry Pi or similar board.

Basics of Python and C programming languages