Skip to main content
Disciplines
Automotive Business IT + Informatics Construction + Real Estate Electrical Engineering + Electronics Energy + Sustainability Insurance + Risk Finance + Banking Management + Leadership Marketing + Sales Mechanical Engineering + Materials
Events
DE
EN
Books
Journals
Topic Page
Marketing
Start single access now
Access for companies
EXTENDED SEARCH
Log in
Top
Published in:
Motion Is the Protagonist of Vision Read chapter Read first chapter

2022 | OriginalPaper | Chapter

5. Information-Based Laws of Feature Learning

Authors : Marco Gori, Alessandro Betti, Stefano Melacci

Published in: Deep Learning to See

Publisher: Springer International Publishing

Log in

Activate our intelligent search to find suitable subject content or patents.

search-config
loading …

Abstract

What are the mechanisms behind learning to see? This is what we address this chapter, where the underlying computational process does in fact characterize the agent’s life in his own visual environment. This is built up on the neural architecture described in the previous chapter that is properly chosen for the incorporation of the motion consistent and abstraction constraints coming from the I and the II Principles. We begin addressing the simplest case of feature conjugation that arises when we estimate the optical flow and continue by considering the canonical set of ODE that express all the visual constraints.

Dont have a licence yet? Then find out more about our products and how to get one now:

Springer Professional "Wirtschaft+Technik"

Online-Abonnement

Mit Springer Professional "Wirtschaft+Technik" erhalten Sie Zugriff auf:

  • über 102.000 Bücher
  • über 537 Zeitschriften

aus folgenden Fachgebieten:

  • Automobil + Motoren
  • Bauwesen + Immobilien
  • Business IT + Informatik
  • Elektrotechnik + Elektronik
  • Energie + Nachhaltigkeit
  • Finance + Banking
  • Management + Führung
  • Marketing + Vertrieb
  • Maschinenbau + Werkstoffe
  • Versicherung + Risiko

Jetzt Wissensvorsprung sichern!

inform now

Springer Professional "Technik"

Online-Abonnement

Mit Springer Professional "Technik" erhalten Sie Zugriff auf:

  • über 67.000 Bücher
  • über 390 Zeitschriften

aus folgenden Fachgebieten:

  • Automobil + Motoren
  • Bauwesen + Immobilien
  • Business IT + Informatik
  • Elektrotechnik + Elektronik
  • Energie + Nachhaltigkeit
  • Maschinenbau + Werkstoffe




 

Jetzt Wissensvorsprung sichern!

inform now

Springer Professional "Wirtschaft"

Online-Abonnement

Mit Springer Professional "Wirtschaft" erhalten Sie Zugriff auf:

  • über 67.000 Bücher
  • über 340 Zeitschriften

aus folgenden Fachgebieten:

  • Bauwesen + Immobilien
  • Business IT + Informatik
  • Finance + Banking
  • Management + Führung
  • Marketing + Vertrieb
  • Versicherung + Risiko




Jetzt Wissensvorsprung sichern!

inform now
previous chapter Foveated Neural Networks
next chapter Non-visual Environmental Interactions
Footnotes
1
A soft satisfaction of the \(b \bowtie v=0\) conjugation.
 
2
The aperture problem and the barber’s pole.
 
3
A very nice example is illustrated at http://​elvers.​us/​perception/​aperture/​.
 
4
Here, for simplicity and for not having to deal with too many indices, we are considering images which have only one channel, like grayscale images; however, if we let \(t\mapsto b(t)\in \mathbb {R}^{w\times h\times c}\) where c is the number of channels of the image at time t (e.g., \(c=3\) for RGB images), we recover the general case.
 
5
This can actually be proven assuming regularity on \({\mathrm {HS}}(\cdot ,t)\); in particular, if \({\mathrm {HS}}(\cdot ,t)\in \mathrm{C}^\infty (I\!\!R^N)\), then also \(\omega \in C^\infty ([0,T])\).
 
6
Here, we use Einstein convention and we do not explicitly write the two components of f and \(\mathrm {D}\Phi \) for compactness.
 
7
Here, we are using the shifted brightness \(b(t+\tau )\) instead of the temporal derivative as indicated in Chap. 3 because, while clearly related for suitable choices of \(\tau \), this solution seems to be more straightforward to implement.
 
8
With the usual conventions for the Einstein summation, and where \(|\cdot |\) indicate that we are summing also on the spatial components of A.
 
9
Online learning on video is mostly unexplored!
 
10
Blurring the video for optical flow and beyond.
 
11
Blurring and foveated nets.
 
12
Feynman’s discovery of Bessel’s equation in capacitors and blurring processes.
 
Metadata
Title
Information-Based Laws of Feature Learning
Authors
Marco Gori
Alessandro Betti
Stefano Melacci
Copyright Year
2022
Book
Deep Learning to See

Print ISBN: 978-3-030-90986-4

Electronic ISBN: 978-3-030-90987-1

Copyright Year: 2022

DOI
https://doi.org/10.1007/978-3-030-90987-1_5

Premium Partner

    Image Credits