Computer Vision III: Detektion, Segmentierung und Tracking

What it is about

You will learn modern methods for object recognition, segmentation and tracking in images and videos. In the end you will understand the underlying deep-learning concepts and be able to work on real computer-vision problems with state-of-the-art models in PyTorch.

What you will be able to do

• Understanding proposal-based and one-stage detectors
• Application of instance-, semantic- and panoptic-segmentation
• Methods for video object segmentation and visual object tracking
• Multiple object tracking and trajectory prediction
• Use of graph neural networks and 3D-tracking approaches

What the module consists of

• VorlesungTheoretical background on neural networks and deep-learning architectures for detection, segmentation and tracking
• Praktische EinheitenHands-on work with state-of-the-art models in PyTorch; central role for understanding and implementation

Teaching method

• Vorlesungexplains the theoretical foundations and models
• Praktische Übungen/Projektestudents get familiar with implementation, training and optimization