DeckVision

Build better card detection models with real-world data. DeckVision auto-captures and labels card layouts so you can focus on improving accuracy.

DeckVision turns your iPhone into a playing card training data collector. Point the camera at any arrangement of cards, and DeckVision automatically detects and labels each one in real time using on-device machine learning. When detections stabilize, the image is captured automatically -- no tapping required.

Every captured image goes into a review queue where you can inspect bounding boxes, correct misidentified cards, and delete false positives before they pollute your dataset. Edits are non-destructive, and you can return to any image at any time.

When your dataset is ready, export everything as a single zip file in Create ML object detection format. The archive includes all annotated images, an annotations.json file with bounding boxes and card labels, and a metadata.json file with detection confidence scores and correction history for full provenance tracking.

Key features

• Real-time card detection with automatic capture when the scene stabilizes
• Full annotation review and correction after each scan
• Browsable image library with thumbnail grid and detail editing
• Create ML-compatible export with annotations and metadata
• 100% on-device processing -- no data ever leaves your iPhone
• Camera permission managed transparently, revocable at any time

DeckVision is built for machine learning practitioners, researchers, and anyone working to improve playing card detection models with real-world, human-verified training data.