Why One Desk Photo Should Give You 12 Products, Not 1
Point a visual search tool at a desk setup and ask "what is this lamp?" and you'll usually get an answer. Ask about the keyboard next to it, and you have to start over -- crop, re-upload, ask again. A desk photo doesn't have one subject. It has a keyboard, a mouse, a monitor, a lamp, a mouse pad, maybe a plant and a laptop stand -- ten or more distinct products sharing one frame. Most tools are built to answer "what is the one thing in this photo," not "what are all the things in this photo." That gap is the whole reason a second desk-setup post is worth writing.
The single-item assumption is built in, not incidental
This isn't a small gap in an otherwise similar tool -- it's a different question being asked. A model built to answer "identify this object" naturally converges on the one dominant subject in frame: the thing that's largest, most centered, or most visually distinct. That's the right behavior for a photo of one product. It's the wrong behavior for a desk, a room, or an outfit, where the "one dominant subject" framing throws away everything else in the shot.
Zomony's vision pipeline asks a different question from the start: not "what is this," but "how many distinct, purchasable products are visible in this image, and where is each one." The prompt sent to the model (Claude, via Amazon Bedrock) explicitly asks for every product it can find, each with its own bounding box -- not a single best-guess label for the scene. See how Zomony finds every product in a photo for the mechanics.
What this looks like on an actual desk
Upload one photo of a real desk setup and you don't get "a desk." You get a keyboard, a monitor, a desk lamp, a mouse pad, and whatever else is genuinely in frame -- each one boxed, labeled, and numbered on the photo, each with its own real search link. One upload, one wait, the whole desk instead of one item at a time.
That matters most for the way people actually encounter desk setups: not standing in front of their own desk with one specific question, but scrolling a "desk tour" video or a coworker's photo and wanting to know about several things in it at once -- the lamp and the monitor arm and whatever that little stand is. Re-running a single-item search once per object in the photo is exactly the friction that makes people give up and never actually buy anything.
Why this is a design decision, not a feature checkbox
It would be easy to describe this as "Zomony finds multiple products," as if it's one extra feature bolted onto single-item recognition. It isn't -- it's a different starting question, and it shapes everything downstream: how the prompt is written, how confidence scoring works when objects overlap or sit close together, how many results is "too many" before the photo overlay gets unreadable (Zomony caps this at the 10-15 highest-confidence detections per photo, tuned specifically for cluttered scenes like a desk). None of that exists if the underlying question is still "what is the one thing here."
What's still honestly limited
Each detected item gets a real, live search link to Shopee or Lazada -- not a fabricated listing. The product data behind that link (title, price, image) is still mocked while real per-marketplace catalog search is built out; see how Zomony finds every product in a photo for the full, current picture of what's real versus what's still ahead. We'd rather say that plainly than imply more than what's actually built today.
Try it on a desk with a lot going on
The more crowded the desk, the more this difference shows up -- a single clean product photo doesn't really test it. If you've got a "desk tour" screenshot or your own genuinely cluttered setup sitting in your camera roll, that's the fastest way to see it: upload it and count how many separate items come back. The same detection works on any photo, not just desks, and it's free to try.