Copyright – use of images to train algorithm

Question

I am working on a university project about computer vision. Part of my work requires object recognition, for which I would use cascade classifiers. These cascade classifiers need to be “taught” what an objects looks like and therefore I need a big collection (hundreds) of images. The problem is, that I cannot create enough material on my own and need to use images from the Internet.

This would usually not be a big issue, because there are a lot of academic sources for computer vision and machine learning, as well as many places for public domain and Creative Commons images, but in this case it is a little difficult.

I need photos of bird feces, which are not available in a collected manner.

On aggregated licensed data, I would cite the source and apply additional conditions if a license requests to do so. The images do not get (re-)published, but they are used to generate classifiers that the object-detection algorithm uses. I do not want to infringe copyright.

Fair use does not apply as I am from Germany. How should I optimally proceed to be on the save side?

Answer 1

I quote from the German copyright (Urheberrecht, original, English translation):

§ 15 Allgemeines

(1) Der Urheber hat das ausschließliche Recht, sein Werk in körperlicher Form zu verwerten; das Recht umfaßt insbesondere

1. das Vervielfältigungsrecht (§ 16),
2. das Verbreitungsrecht (§ 17),
3. das Ausstellungsrecht (§ 18).

(2) Der Urheber hat ferner das ausschließliche Recht, sein Werk in unkörperlicher Form öffentlich wiederzugeben (Recht der öffentlichen Wiedergabe). […]

---

Article 15 – General

(1) The author has the exclusive right to exploit his work in material form; this right shall in particular include

1. the right of reproduction (Article 16),

2. the right of distribution (Article 17),

3. the right of exhibition (Article 18).

(2) The author further has the exclusive right to communicate his work to the public in non-material form (right of communication to the public). […]

This paragraph tells you what is protected by German copyright in the first place and the only thing that you need to do for your endeavour seems to be copying/reproducing the photos, as you do not seem to distribute, exhibit or communicate them.

§ 53 Vervielfältigungen zum privaten und sonstigen eigenen Gebrauch

[…]

(2) Zulässig ist, einzelne Vervielfältigungsstücke eines Werkes herzustellen […]

1. zum eigenen wissenschaftlichen Gebrauch, wenn und soweit die Vervielfältigung zu diesem Zweck geboten ist und sie keinen gewerblichen Zwecken dient,

2. […]

---

Article 53 – Reproduction for private and other personal uses

[…]

(2) It shall be permissible to make single copies of a work […]

1. for one’s own scientific use if and to the extent that such reproduction is necessary for the purpose and it does not serve a commercial purpose,
2. […]

This paragraph tell you that the only thing that you need to do which might infringe copyright as assumed above (reproduction) is permissable if you do it for your own scientific use and if it is necessary (which seems to be the case).

How should I optimally proceed to be on the save side?

Check the entirety of the above paragraphs and the rest of the German copyright to ensure yourself whether there is anything that would make your endeavour illegal.

Also check whether other copyrights may apply. The Berne Convention may ease things a lot.