Deep Learning on Image and Signal

The aim of the module is to teach advanced techniques of image and signal processing using deep learning. Students will gain insight into the fascinating performance of modern image and signal processing systems. The methods covered build on techniques from stochastics, analysis, machine learning and deep learning.

Table of contents

• Data preparation of the images and signals for deep learning
• Practical considerations when applying deep learning models
• Object recognition and segmentation in image data using deep learning
• Prediction of signals (audio, sensor data) using deep learning

Students will be able to prepare signals and images for use in deep learning models. This includes signal and image transformation, data cleaning, normalization, centralization or data augmentation, generation of ground truth data. In addition, the students know how to apply best practices for model training using deep learning for signals and images.

Students will be able to successfully use deep learning models for object recognition in image data such as faster R-CNN or YOLO and adapt them to application needs.

Students will be able to implement and use image segmentation using encoder-decoder deep learning models such as U-Net or Auto-Encoder.

Students will be able to implement and use deep learning models such as RNNs or LSTMs to predict signal data (audio, sensor data).