Hi everyone,
I’m performing an fMRI experiment on two categories of picture stimuli (i.e. animals and tools). It’s more difficult for participants to distinguish among members of one category. I’m interested in rating my fmri stimuli based on their within category visual similarity. Is there any data set that attributes visual similarity scores to subcategories of objects (e.g. cars, birds …)? I would also appreciate any other suggestion for rating each picture based on within category visual similarity.
Thanks in advance
Haleh. Kh. D
Hi Haleh, I cannot think of any existing picture norms have that information. The closest one that I think still quite helpful in your case comes from Cree and McRae (2003) and their following norm publication in 2005. In their study, they are people to write down features associated with >500 concepts across many categories. Based on the featuring listing data, they calculated the visual similarity measures for each subcategory (e.g., fruits, weapons, vehicles, birds, animals, etc.). Although these data comes from verbal modality, I think they may help you get a rough sense of the visual similarity for the subcategories.
For collecting picture rating data, in order to get similarity data, the traditional method is to collect all pair-wise comparisons and calculated the rated similarity scores for each subcategory. This will be painful. I will suggest you using the adaptive sampling algorithm which is powerful and easy to use. I have add the links for relevant paper and website for the adaptive sampling. Hopefully, it will be helpful! Good luck!
http://link.springer.com/article/10.3758/BF03192726#page-1
http://psycnet.apa.org/psycinfo/2003-03979-003
http://homepages.cae.wisc.edu/~jamieson/resources/activeRanking_extended.pdf
https://next.discovery.wisc.edu/
I think there are some subtests on certain intelligence tests that test the individual's ability to identify visual similarities. One that I can think of is the Wechsler symbol search tests. I'm not sure that this one fits your criteria and I can't off the top of my head think of which tests measure this.
dear haleh,
depending on your research question you might want to consider using morphed stimuli to parametrically vary the inter-stimulus similarity (eg http://www.ncbi.nlm.nih.gov/pubmed/20362595).
good luck!
This is a new answer to an old question: We have a small image pair similarity dataset: imagesim-353. I have pointed to it here: https://finnaarupnielsen.wordpress.com/2020/07/27/imagesim-353-image-pair-similarity-dataset-for-deep-learning-and-knowledge-graphs/ and the direct link is here: https://people.compute.dtu.dk/faan/data/imagesim-353.zip
We used it for testing a model described in "Inferring visual semantic similarity with deep learning and Wikidata: Introducing imagesim-353 http://www2.imm.dtu.dk/pubdb/edoc/imm7102.pdf
I am not sure our dataset is useful for you, and it is probably also late for you.
- Badges
- Science topic