**The course is taught by Mirosław Korzeniowski from Tooploox.**
In this lecture we will go over techniques and applications of Convolutional Neural Networks (CNNs) which constitute a large branch of Deep Learning. We will illustrate them with practical examples using TensorFlow and Keras.
We will start from introducing convolutions based on their applications in traditional Computer Vision and follow the evolution of Deep Learning techniques based on convolutions. We will see how one can adapt existing convolutional networks architectures to their own problems through changing parameters of convolutions and other layers, choosing proper activation functions in both hidden and final layers, as well as playing with metrics and loss functions.
As examples we will mainly choose 2D images and videos but we will also look at 3D applications, which are popular especially in medical image processing. We will also see how these image processing techniques can be applied to 1D data such as analysis of sound, power consumption, and GPS tracking.
The lecture will be accompanied / interleaved with labs where students will work on practical projects in small groups.
