Slides:

• https://www.notion.so/Week-1-Introduction-Pinhole-Model-be8ca5f9d01f4e3388bd06af39ec5e48?pvs=4#c50ff12a6ab84b7ba9c7c3d8addebeae
• https://www.notion.so/Week-1-Introduction-Pinhole-Model-be8ca5f9d01f4e3388bd06af39ec5e48?pvs=4#906df5d739974097bdb28954b16a6a3e

(Wedesday 2023-09-20)

Introduction

By Marc Pollefeys and Siyu Tang. Slides from Luc Van Gool, Andreas Geiger, Ioannis Gkioulekas, and many others.

What is computer vision?

What a human sees: looking at it, you automatically start recognizing objects, imagine the place, etc.

What a computer sees: a matrix of numbers, encoding the amount of light in that location (or for the 3 colors…)

The goal of computer vision is to give computers (super) human-level perception (in a general setting we are very far from that, but in some narrow tasks cv algorithms can be already better than human level)

The goals more concretely: automatic understanding of images and video

• Computing properties of the 3D world from visual data (measurement, reconstruction)
• Algorithms and representations allow a machine to recognize objects, people, scenes, and activities. (perception and interpretation)

Vision for measurement examples:

• Mars rovers need to map their surroundings to be able to choose their next moves (bc information exchange is ≥ 8 mins, so a certain autonomy is needed) → stereo camera system

  

• 3D surface models of buildings from videos or set of photos (see prof. Pollefeys’ PhD thesis: https://people.inf.ethz.ch/marc.pollefeys/pubs/PollefeysPhD.pdf) (image on the right)

• 3D models from photo collections from the internet, using better feature algorithms

Vision for perception, interpretation:

• to find objects, activities, scenes, locations, text, faces, gestures, motions, emotions,…

What does it mean, to see? (asked in the very influential book: Vision by David Marr)

• The plain man’s answer (and Aristotle’s, too) would be, to know what is where by looking.
• To discover from images what is present in the world, where things are, what actions are taking place, to predict and anticipate events in the world (predict the future).

Relation to other disciplines

Related disciplines:

• algorithms (large amounts of data → algos have to be efficient, otherwise it won’t be ready for the conference deadline)