I am creating a webtool for image moderation (approve / flag / delete). My starting image set to review is about 200,000 with 400 images added per hour.

The end users of this tool will be outsource workers who will both clear through the backlog and continue to review the incoming images.

My prototype displays batches of a grid of 4x4 images of 240x240.

What is the ideal image size for quick and easy review (e.g. nude / not nude) and what is the ideal grid size for each batch?


Isn't this the perfect thing you can test for? Make some wireframes and test. Make some higher fidelity prototypes and test again. This is not something you have to guess about, nor does it make sense to ask others to guess for you.

On a side note - if you are testing for nude/not nude then I think you can go a lot smaller than 240x240.

I just went to your network profile and saw that you posted mostly in network engineering. The best steps would be to go to UX/IA person (or designer or front end developer if UX people aren't available) and have them create a wireframe (use real photos in this wireframe). The developer can make a variety of grids and resize the photos in the code (this is a test after all and not production). Then ask people. See what works and see what doesn't.

  • You are correct that we can test for this stuff but as you said, I don't have a lot of experience in UI/UX and was hoping there was some experience I could learn from. This is also why I posted in a UX forum and not in a GUI forum. – SAR622 Oct 28 '14 at 14:10
  • That's why I modified my comment at the end. : -) Get a developer to make a prototype. What screen size are the testers using? 1280? Make v1 with 240x240 as you suggested (you can do 5 across). Then do a v2 with a 150x150 and another w 120x120 and another even smaller. Then look at it yourself. Ask other people in your team: sys admins,PMs, business developers, your wives, brothers, friends and see what they respond to. Best of all is asking the people who will be doing the work. You'll quickly find what works best. In fact this is probably the ideal first project. – Mayo Oct 28 '14 at 14:18
  • Testing, at it's most basic level is asking people: do you like this one? or this one? – Mayo Oct 28 '14 at 14:20

While this is not an absolute answer, here is what I found after some testing:

I used a grid of 6x10 with each image at 180x180px. My data set had an estimated percentage of 1% of the images needing to be flagged. By setting all images to a default of approved and requiring interaction only on the images to be flagged, I was able to do 4 sets in a minute.

Assuming 70% work time over an hour (I didn't do the exact cycle and takt time calculations) this meant I was able to expect 10,000 images in an hour.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.