axiom-ios-vision_skill

Overview

This skill provides a single, battle-tested router for any iOS computer vision work that uses the Vision framework. It directs you to implementation patterns, API references, or diagnostics depending on whether you are building features, looking up APIs, or debugging detection issues. Use it for image analysis, detection, segmentation, OCR, barcodes, and live scanning patterns across xOS platforms.

How this skill works

The router inspects your intent (pose, segmentation, OCR, barcodes, documents, live scanner) and forwards you to one of three focused areas: implementation patterns, a complete Vision API reference, or diagnostics. Implementation patterns contain ready-to-use examples (hand/body pose, subject lifting, person segmentation, text and barcode workflows). The API reference lists key VNRequest classes, DataScanner integration, and coordinate conversion patterns. Diagnostics covers common failure modes and performance tips.

When to use it

• Analyzing images or video streams with Vision requests
• Detecting hands, bodies, faces, or tracked landmarks
• Segmenting people or isolating a subject from background
• Reading text (OCR), scanning barcodes/QR codes, or scanning documents
• Implementing live scanning using DataScannerViewController
• Troubleshooting detection failures, low confidence, or coordinate bugs

Best practices

• Choose the right request and confidence thresholds for accuracy vs performance
• Use coordinate conversion helpers between Vision and UIKit/Metal/CoreImage
• Prefer instance masks and hand-exclusion patterns when isolating subjects
• Use fast vs accurate VNRecognizeTextRequest modes depending on latency needs
• Select barcode symbologies to improve detection reliability
• Profile performance on device and limit request frequency for realtime streams

Example use cases

• Detect 21-point hand pose in a camera frame and track gestures
• Segment a photographed person and composite a new background using HDR safe CoreImage blending
• Implement live OCR overlay with DataScannerViewController for instant text capture
• Scan documents with perspective correction using VNDocumentCameraViewController
• Detect QR codes with selected symbologies and handle scan distance heuristics