paper_scanner_ios
The iOS implementation of paper_scanner, using Apple Vision and
CoreImage.
Apps should depend on paper_scanner (this package is pulled in automatically
as the endorsed iOS implementation).
How it works
| Operation | API |
|---|---|
| Detect | VNDetectDocumentSegmentationRequest (iOS 15+), fallback VNDetectRectanglesRequest |
| Crop | CoreImage CIPerspectiveCorrection |
| Filter | CIColorControls (enhance), CIPhotoEffectMono (grayscale), CIColorThreshold/mono+contrast (black-and-white) |
The document-segmentation request is runtime-gated to iOS 15+ and gracefully
falls back to rectangle detection on older systems. Vision coordinates use a
bottom-left origin and are flipped to the top-left origin used across the
plugin.
Why Vision and not VisionKit?
VisionKit's VNDocumentCameraViewController renders its own camera, crop
and review UI (private ICDocCam* classes) and exposes only a results
delegate — there is no way to restyle it. To build a custom UI we use the
detection-only Vision requests instead.
Requirements
- iOS 13.0+ (
platform :ios, '13.0'). - Host app must declare
NSCameraUsageDescriptioninInfo.plist.
Libraries
- paper_scanner_ios
- The iOS implementation of
paper_scanner.