vision-framework

To add OCR and barcode scanning to an iOS camera feed, use VisionKit DataScannerViewController for live scanning or process video frames with Vision requests. This skill extracts text and barcodes, returning results as bounding boxes with confidence filtering.

The Apple Vision framework supports both modern iOS 18+ Swift-native async request APIs and legacy VNRequest approaches. This skill prioritizes the modern async layer while documenting the legacy method to ensure compatibility across different iOS API generations.

Yes, you can detect faces, landmarks, and capture quality, plus track objects across video frames using Vision. This skill applies tracking capabilities to real-time video feeds, allowing continuous object recognition and facial feature extraction within iOS applications.

To handle Vision's normalized coordinate system safely, apply proper coordinate transformations when mapping detected bounding boxes to your UI. This skill ensures correct runtime behavior by implementing safe coordinate handling alongside VisionKit availability checks.

Structured document recognition requires iOS 26+ Vision APIs, while optional Core ML inference is supported through custom Vision requests. You do not strictly need Core ML for standard OCR, but it enables extended inference for specialized models.

VisionKit DataScannerViewController fails on older iOS versions when availability checks are not performed before runtime. This skill implements VisionKit availability checks to ensure correct runtime behavior and prevent crashes on unsupported iOS versions.