Tailored for multiple docs on a single page. Isolate and categorize each document.

Edge detector feature is a specialized computer vision feature that detects, isolates, and crops individual records scanned together on a single page.

Instead of treating a scan of several receipts or IDs as one image, the API identifies the physical boundaries of each item. It then creates separate, high-resolution standalone files for each record, ensuring that the downstream extraction engine processes them individually for maximum accuracy.

Edge detector finds the borders of each document within a single image or scan and separates it, so each one is processed as its own record. Common examples:

Expense management: An employee scans three or four taxi receipts and a restaurant bill on one page. Each receipt is cropped out and extracted separately, so a lunch receipt is never merged with a taxi fare.

Identity verification (KYC): The front and back of a driver’s license, or several ID cards, are scanned together. Each side or card is isolated for verification.

Healthcare: Several lab labels or prescription stickers are scanned onto one sheet. Each one is cropped and digitized individually.

See more examples in our customer stories.

Mindee's process is a sophisticated blend of geometric analysis and edge detection. When a single page is uploaded, Mindee’s vision pipeline performs a "spatial scan" to find high-contrast edges that indicate a physical object's edge.

• Object detection: The AI identifies how many distinct items are on the page.
• Bounding box generation: It draws a precise "box" around each item, even if they are placed at an angle (this is where "auto-deskewing" happens).
• Cropping & de-noising: The system crops each box into a new, individual image file and clean the quality.
• Individual processing: Each new file is sent to the relevant extraction model (e.g., the Receipt API) as if it had been uploaded on its own, eliminating cross-talk between different documents.

Mindee’s edge detector feature does this automatically. When you send an image or PDF containing several documents (for example, a photo of multiple receipts on a table or a scan of several IDs), it detects the borders of each document and returns its exact position, so each one can be cropped out and processed individually.

You don’t need to crop files by hand or write your own image-processing code. You can accept bulk scans from users and let the API handle the separation, which saves development time and makes uploading easier for your users.