Intro

Visual Intelligence (VI) — the next generation of Computer Vision — is transforming how machines perceive, interpret, and interact with the world. Fueled by deep learning breakthroughs, VI powers applications ranging from autonomous driving and medical image analysis to creative AI systems that generate images, videos, and 3D content. With rapid advances in areas such as Vision Transformers, Diffusion Models, and Large Multimodal Models, this is a field evolving at unprecedented speed. Studying VI now means positioning yourself at the forefront of AI innovation, where visual understanding is a cornerstone of intelligent systems.

Content

The course builds from the fundamentals of deep learning to advanced state-of-the-art methods in modern computer vision. Topics include image classification, object detection and tracking, segmentation, pose estimation, depth prediction, generative and multimodal AI (GANs, diffusion models, vision-language systems), and 3D reconstruction (e.g., NeRFs and 3DGS). We focus on the key concepts from the deep learning side of computer vision alongside recent breakthroughs, supported by hands-on tasks and discussion of both seminal and cutting-edge research papers.

Tasks and Evaluation

The evaluation in this module will be based on some smaller tasks — including a Generative AI–based Educational Challenge — and a more practical, real-world mini project. You will have the freedom to somewhat focus on topics that inspire you and align with your master’s thesis, while deciding how to balance deep specialization with broader exploration. Through sharing and discussing results, you will also benefit from your peers’ work and gain valuable insights into the wider field of Visual Intelligence.

Teacher: Frank Lindseth

Sign up:

More info and how to sign up for non-IDI students can be found here:

Info, Modules, TDT4506 (Computer Science), IT3020 (Informatics), Registration (other study programmes)

Background

The background of the students interested in this module is probably a bit broad.  Its an advantage that we are somewhat on the same page when the module starts. If you know that you are a bit rusty on things like calculus/partial derivatives/lin.algebra, Python/NumPy/PyTorch/Notebooks and DL/NNs please try to read up a bit, a lot of resources are availble online and is only a search away. As for the DL/NN part a gentle intro can be:

• Neural Networks (NN, 3Blue1Brown)
• Neural Networks and Deep Learning (Nielsen)