For the complete documentation index, see llms.txt. This page is also available as Markdown.

Crop Result

Reference documentation on processing a Crop result using the Mindee SDKs.

You'll need to have a response object as described in the Response Processing section.

Accessing Crop Items

A CropItem describes the location of a single detected object.

In this context, an object can be any element to find on a page.

Most users will look for a document type (like a receipt or an ID), but it could really be anything (a photo, a logo, etc).

CropItem Attributes

Object Type

The category assigned to the object. It is always filled.

It is returned as a string and is identical to the value entered on the Mindee platform — case, spaces, and punctuation included.

Why "Object Type" instead of "Document Type", like for Split and Classification? Because the fundamental technology is different, Crop uses Object Detection algorithms, whereas Split and Classification use variations of classification technology.

Location

The location of the object in the document. It contains the following properties:

  • Polygon: Coordinates of the object.

  • Page: 0-based index of the page the coordinates were found on.

Extraction Response

Optional extraction response associated with the split. This is only filled if extraction chaining is activated for the model.

Iterate Over Crop Items

You'll usually want to iterate over all crop items, since the number of items is dependent on the document. Remember that a document can have multiple pages, and each of its pages can have multiple crop items.

Chained Extraction results will be an ExtractionResponse object. This object is identical to the response of an Extraction Model.

Refer to the Extraction Result section for details on processing the Extraction.

Last updated

Was this helpful?