LEARNING OUTCOMES

Learning how to represent image content adaptively by means of shallow or deep computational models and biologically-inspired hierarchical models, and how to tackle image classification and categorization problems.

AIMS AND LEARNING OUTCOMES

Students will be provided with an an overview of state-of-the-art methods for modeling and understanding the semantics of a scene. Students will get acquainted with the problem of representing the image content adaptively by means of shallow or deep computational models. Then it will address image classification and categorization problems. Possible extensions to depth and motion information will also be discussed.

Students will be involved in project activities.

PREREQUISITES

Calculus and linear algebra.

Digital image processing and machine learning principles.

TEACHING METHODS

Theoretical classes complemented by practical activities

SYLLABUS/CONTENT

Course content

• Elements of classical Computational Vision
  • Review of image processing: image  filtering, feature detection, ...
  • Image matching: feature detection, description and feature similarity between image pairs
  • Multi-scale and  multi-resolution representations
  • Motion analysis  and optical flow
• Computational Vision and Machine Learning algorithms
  • Bag-of-words representations and image classification
  • Sparse coding on fixed over-complete dictionaries: application face detection
  • Unsupervised segmentation and super-pixel computation
• ​Computational Vision and Deep Learning:
  • Principles of Deep Learning e Convolutional Neural Networks
  • Convolutional methods for multi-object detection
  • GANs: principles and applications
• Projects and use cases

RECOMMENDED READING/BIBLIOGRAPHY

material provided by the instructors (slides and papers), see course Aulaweb page

additional reference online book http://szeliski.org/Book/