Ask HN: Is this plagiarism?

[hojijoji]

https://www.cv-foundation.org/openaccess/content_cvpr_2015/papers/Hofmanninger_Mapping_Visual_Features_2015_CVPR_paper.pdf

vs.

https://pdfs.semanticscholar.org/3276/436e8ceffa3f2ed1e4b648fbef7c85b59616.pdf

[davelnewton]

Except for the obvious image copy without attribution, it doesn't seem to be. At first blush it looks to be a specific application of the previous paper's technique(s).

Why?

[hojijoji]

There is no cite, they seem to change the wordings of the paper to avoid 1 to 1 copy of text. E.g. they change 'label distribution model' with 'lesion distribution model'. Their chapter 'Learning Image-Text Mapping Model' basically describes the method of the first paper with minor but bad changes and it is not really self contained. They claim they simply forgot to cite. For me, it doesn't look like forgetting a cite. I wanted to hear different opinions.

[Someone]

Seems like davelnewton gave you a different opinion.

[hojijoji]

What is your opinion? Is it only plagiarism if its a copy? Can something be 25% plagiarism?

[Someone]

I haven’t looked deeply into it, but I see a paper about a method and one about using a method that may or may not be that same method for detecting aneurysms (I couldn’t find ‘aneurysm’ in the first paper)

The two papers share (part of) a picture that may or may not have been original in either paper.

I haven’t checked which paper was published earlier, but that likely is the second (it cites a 2017 paper, while the first is from a 2015 conference)