Convolutional Neural Networks and Image Processing II

Convolutional Neural Networks and Image Processing II


Jiří Materna


  • basic knowledge of programing in Python
  • high school level of mathematics
  • Basics of machine learning on the level of our course Introduction to machine Learning
  • Knowledge on the level of our basic Convolutional Networks and Image Processing


This course is a follow-up of Convolutional Networks and Image Processing in which we will focus on image data preprocessing and advanced techniques of deep learning for image processing. Apart from image classification, well known from the previous course, we will study image segmentation, object detection, and especially advanced applications of generative adversarial networks (GANs) such as superresolution, noise reduction and generating deep fakes.


  • Image data preprocessing (number of channels, aspect ratio, image scaling, normalization of image data)
  • Data augmentation (scaling, rotation, shifting, mirroring)
  • Practical example of preprocessing and augmentation
  • DNN architectures for Image Processing (convolution, deconvolution, pooling, residual) connections
  • Advanced image classification (ImageNet dataset, VGG, ResNet, Inception, MobileNet, Efficient net)
  • Practical example of transfer learning using different base models
  • Image Segmentation (U-net, Object detection)
  • Practical example of image segmentation and object detection
  • Generative Adversarial Networks (generative networks, sampling, the adversarial model)
  • Practical example of image generation
  • Image Noise Reduction
  • Practical example of noise reduction using U-net and GANs
  • Superresolution (Upsampling, practical example of using GANs for superresolution)
  • GAN enhancement of datasets
  • Deep fake (Conditional GANs, examples of deep fake techniques)
  • Adversarial attacks (image modification, adversarial patch, black box-attacks, cybersecurity)


If you wish to enroll in this course please contact us on