Run Inference (v1 - Latest)

Decide whether an item is damaged, possibly damaged or not damaged from 1 to 10 photos, with the overall severity and each damaged area's type, severity, confidence and bounding boxes. Every photo is checked first. If any photo is rejected (too small, blank, too dark, overexposed, blurry, not a real photo, obstructed, showing several items, or not showing the named item), damage is not evaluated and the request responds with HTTP 400, listing each rejected photo's image_index and reason_code. Submit an image to analyze using the Product Damage Detector agent.

Recent Requests
Log in to see full request history
TimeStatusUser Agent
Retrieving recent requests…
LoadingLoading…
Body Params
files
array of objects
required
length between 1 and 10

1 to 10 files, in the order you want them numbered.

Files*
string
length ≤ 200

Optional. What the item is, such as 'baby stroller' or 'lipstick'. When set, photos that do not show this item are rejected with reason_code item_not_found, and damage is judged for this kind of item.

string
length ≤ 1000

Optional. Your own rules for what counts as damage for this item, such as 'scuffs on the outer box are not damage'. Applied on top of the standard label definitions.

Headers
string
required

Your API key for authentication

Responses

Language
LoadingLoading…
Response
Click Try It! to start a request and see the response here! Or choose an example:
application/json