Mitchell van Zuijlen: 19000 annotated paintings that can be interdisciplinary used

The 4TU.ResearchData dataset from Mitchell’s Ph.D. project got down­loaded over 1200 times in 28 coun­tries only 1.5 years after upload­ing it. “When I was pre­sent­ing my research at con­fer­ences, peo­ple were often pos­i­tive about the idea of pub­lish­ing the data open access,” Mitchell says. “How­ev­er, you nev­er know what will hap­pen when you pub­lish the data. It was a big relief to final­ly put it online and very cool to see all these peo­ple using it now.”

Mitchell and his col­leagues col­lect­ed more than 19000 open source paint­ings from insti­tutes like gal­leries, muse­ums, libraries, and archives, and cre­at­ed an anno­tat­ed dataset from it. “We used our dataset to study the depic­tion and per­cep­tion of mate­ri­als through the artists’ eye”, explains Mitchell. “We always had the idea that if the data are inter­est­ing to us, they might be inter­est­ing to oth­ers as well. That’s why we decid­ed to make it open and acces­si­ble to every­one. The most amaz­ing and excit­ing thing about open­ing up our data has been that peo­ple are using our dataset for pur­pos­es that we haven’t even thought of before.”

Mitchell helps oth­er researchers in his lab now when it comes to the tech­ni­cal skills need­ed to cre­ate and main­tain such an open dataset. “There was no real pro­to­col for get­ting sup­port when I start­ed, which was a bit frus­trat­ing.” He believes that the Open Sci­ence com­mu­ni­ty, includ­ing the 4TU.ResearchData and its sup­port, is becom­ing more and more con­sis­tent, which will save new researchers time and frus­tra­tion.

“My next step for this spe­cif­ic dataset is to add more or more spe­cif­ic cat­e­gories to the data. In the future it would be great to cre­ate a dataset where every­one can add and edit data,” Mitchell con­cludes. “I want to con­tin­ue cre­at­ing Open Sci­ence when­ev­er pos­si­ble.”

Links to Mitchell’s data:

• Dataset: Mate­ri­als In Paint­ings (MIP): An inter­dis­ci­pli­nary dataset for per­cep­tion, art his­to­ry, and com­put­er vision (https://doi.org/10.4121/13679200.v1)
• Web­site: Mate­ri­als in Paint­ings Dataset (https://materialsinpaintings.tudelft.nl/)

Con­tact Mitchell @mitchellvanzuijlen@gmail.com.

Post writ­ten by Tonke L. de Jong.