There are some topics which are inherently good for publications and citations. If you need a specific dataset for your study, this is quite often very difficult to build up. So if one does the work, a lot of people will use and cite your dataset. Even more so, if there are only limited ways to gather such a dataset, especially if you need patients with a specific disease or something like that. In these cases the academia is happy for any dataset they can have, and therefore it gets easily published.
If you look at the published papers, there will certainly be some paper on "here we made this dataset, feel free to use it and cite us", then after this comes the "we now evaluated this dataset with our methodology" sometimes followed by "with this methodolody we made this workflow and product" and so on. Sometimes accompanied with some more datasets, and then you have 10 publications in short time. Each of them building on a reputable and much cited dataset. So these paper gather good traction also.
I was at conferences where half the people just presented new datasets of gazes, pupils, fingerprints, smiles whatever. A lot of tedious work to build up, but it may be a good thing to build up reputation.
Sadly in the current times, reputation in numbers (citations, papers, etc) are sometimes regarded more worth then the content of these numbers.