I am writing up some research I did using a Kaggle dataset: https://www.kaggle.com/c/dogs-vs-cats

I want to include one or two images from the dataset in my paper to show the types of images I'm using. Will this run afoul of copyright issues?

The rules for the dataset say:

These images have been published by Microsoft Research for the express purpose of furthering academic research. They may be used for non-commercial research purposes, but they may not be re-published without the express permission of Microsoft Research.

Including an image in my paper is use for non-commercial research purposes, but it is also re-publishing them without the express permission of Microsoft Research. I asked this question on the Kaggle forum, with no response in several weeks, and there is no contact information for anyone at Microsoft Research.

More generally, is it acceptable to publish examples of images from publicly available datasets even if there is not an explicit copyright license? What about ImageNet, where they took images from all over the internet? Is any paper that publishes examples from ImageNet (there are plenty!) breaking copyright law?

  • 1
    There is always an implicit copyright, even for 'publicly available' images. You need explicit permission either through the license (e.g. like StackExchange's license on content) or from the copyright holder.
    – Jon Custer
    May 30, 2018 at 15:31

1 Answer 1


If you are publishing (like in a journal) then it is for commercial use, just not for your commercial use - the publishing company makes money off of your publication. Therefore you need to get permission as stated in the agreement.

It is almost always a requirement that you get permission under these circumstances. When it is not is when the licence explicitly says so and you are following the rules

  • Ok, but let's assume it is a non-commercial publication (e.g., ArXiv, or a conference run by a non-profit such as the Computer Vision Foundation). Does publishing an example image in a paper count as "research purposes" (as opposed to "re-publishing" the entire dataset)?
    – Katie R
    May 30, 2018 at 16:14
  • 1
    As Jon Custer mentioned, this still counts as re-publishing. What the licence appears to mean is that you may take the images and use them for your research, but can't publish them without permission. What I often see in papers is things like "We used the XYZ data set (see [1]), and then the user just goes to the reference [1] and picks up the data set from there May 31, 2018 at 18:58

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .