Computer vision and machine learning for archaeology - PowerPoint PPT Presentation

1 / 24
About This Presentation
Title:

Computer vision and machine learning for archaeology

Description:

Computer vision and machine learning for archaeology. The potential of machine learning and image analysis techniques for the domain of archaeology ... – PowerPoint PPT presentation

Number of Views:558
Avg rating:3.0/5.0
Slides: 25
Provided by: referentie
Category:

less

Transcript and Presenter's Notes

Title: Computer vision and machine learning for archaeology


1
Computer vision and machine learning for
archaeology
  • The potential of machine learning and image
    analysis techniques for the domain of archaeology
  • Drs. L. van der Maaten, Drs. P.J. Boon, Dr. A.G.
    Lange, Dr. H. Paijmans, Prof.dr. E.O. Postma
  • Paul Boonp.boon_at_cs.unimaas.nl
  • http//www.referentiecollectie.nl/rich
  • Institute for Knowledge and Agent Technology
  • Maastricht University

2
Outline
  • Introduction
  • Why How
  • Computer vision and machine learning
  • Examples
  • Historical glass, on-line reference collection
  • Modern coins
  • Conclusion
  • Future work

3
Why How
  • Need for on-line reference collections
  • Making a collection digital is not enough
  • Create tools for identifying artifacts
  • Vision and machine learning for finding alike
    objects
  • Web technology for presentation and navigation

4
Computer vision and machine learning
5
Computer vision and machine learning
  1. Object detection
  2. Feature extraction
  3. Classification

6
Computer vision and machine learning
Not a replacement, but a tool!
7
Example
  • Historical glass
  • Content based image retrieval
  • Outer shape contours

8
Object detection
9
Shape context
Translation, scale and rotation invariant





?
r
0
0
r
0
0
?
Histogram construction
10
Classification
11
(No Transcript)
12
(No Transcript)
13
(No Transcript)
14
(No Transcript)
15
(No Transcript)
16
(No Transcript)
17
Example
  • Coins
  • Automatic classification system, using texture
    information
  • Modern coins, but in future early-medieval coins

18
Original
19
Object detection
20
Features
21
Conclusion
  • You can successfully integrate computer vision
    and machine learning with an online collection

22
Future work
  • Improve usefulness
  • Object detection
  • Evaluate technology
  • Other types of material
  • Using outer shape and texture
  • Partial shapes, from sherds?
  • Evaluate typologies or device new ones?

23
Shape map
24
Acknowledgements
  • We thank the Dutch Royal Coin Cabinet for
    providing the medieval coin datasetArent Pol
    specialist on early-medieval coins
Write a Comment
User Comments (0)
About PowerShow.com