FAQ
What is a tag?
A tag is simply a single word or term, or a phrase, that describes the image shown. When you search for things online the words you type in the search bar are basically these tags. The search engine if matching the words you used with the places those words appear. If you type "painting" it will only return results that also have "painting" present -- here "painting" is the tag.
"What if none of the AI terms look accurate to me?"
You can skip verifying any of the terms if none of them look accurate. They are not all going to make sense for you, or for the image, and that's okay. We just want to gauge which of the AI terms best suits the images according to you.
"How many terms should I add after I verify the AI tags?"
In the text box you can add your own tags (words, or phrases) that you believe would accurately describe the image shown. You may add as many, or as few, terms as you want. There is no minimum or limit to the tags you can add, but try to focus on adding what you find useful for finding this image. Imagine you are on the phone with a friend and you both have a stack of images, how would you describe the image you are looking at to a friend in order to get them to pick the same image out? Those are the tags to add.
Why did you select the AI Tagging Models you used?
We selected two different AI tagging models for a reason! We selected the Google Cloud Vision API tagger because it is what Google uses to index images in their image search bar, so this is likely one the taggers you will have encountered most in your day-to-day life. We also selected the Metropolitan Museum of Art's iMet tagger because we did not have enough images to accurately train our own tagging model, and the Met had used the iMet Collection Dataset consisting of 155,531 samples which created 1,103 individual tags. We used this tagging model to see what an art trained tagger would tag our visual arts collections with, even though our collection are different than the Met's.
Questions on the Free Tag Workflow "Tag Images"
"How many terms should I add?"
In the text box you can add your own tags (words, or phrases) that you believe would accurately describe the image shown. You may add as many, or as few, terms as you want. We also ask that you separate each tag (word or phrase) with a comma, as it will make it difficult for our team to tell what you actually meant if you do not separate with commas. We also ask that you do not use hyphens (-), underscores (_), or any other symbols. If you want to type a tag that is a phrase or multiple words (for example: young people) please do so with a space between each word. Remember, we are specifically looking for your language, so please add what you would use to describe the image to a friend, or that you would use to find the image for your own future use.
"Are there specific terms you're looking for?"
We are looking for what aspects of the image you would use to find it again. So if there is a term or a phrase that sparks when you see this image, please add it for us. For the purposes of this project we are not looking for any specific, single tag. However, we are mainly looking for how people would describe the images that are shown. Try to focus on the visual aspects of each image you see and add tags (words, terms, phrases) that describe what you are seeing. You do not need to know details of who someone is, or what specific instrument or building is in an image. Instead focus on how you would generally describe the image.
What if my tag/term/phrase has multiple words? How should I enter it?
As we process tags we will keep an eye out for things like formatting and spelling, but ideally we would request that you enter the tags with spaces between words. For example, if you'd like to add a tag for "Space Race," "African American," or "shadow path" we would prefer them entered in this format instead of mashing together as "spacerace," "africanamerican," or "shadowpath". Also we ask you do not use any characters, such as underscores, dashes or & symbols. Please use spaces or write out things like "and." If you are using accents or umlauts in a term, that is okay and should export fine! When we download the terms added by users we will do simple checks for things like spelling mistakes and any broken characters, and will try out best to fix them. The only tags we will not use in our database are tags that are deemed to be actually wrong by Adler's curator - for instance if someone adds that a photo is from 1950s and it's actually from 1964, we will not include the "1950s" tag.
Why is the retirement rate for this project set so much higher than other projects I've worked on?
Our team is setting the retirement rates significantly higher for this project (40 people will need to classify an image before it is removed from the project) because we are not actually looking for one single answer, or any form of consensus. We are looking to get as many people's voices and choices represented; even if people end up saying the same thing, having a more representative sampling of tags helps us to really gauge what terms/tags are helpful to the public when describing our objects. Because of this goal, we ask for your patience and understanding with this higher retirement limit.
Help the Adler know who participated in "Tag Along with Adler," and how your experience was! Fill out the survey here: https://forms.gle/JZ3fuZhKdvahe5dm7