Demo: Running Vision Models Locally on iPhone

amos_gyamfi · x · 2026-07-15

The author created a locally-run SwiftUI vision demo on an iPhone featuring **image classification, object detection, and instance/semantic segmentation**, emphasizing that all processing occurs entirely on-device. Listed model capabilities include: - Classification: Selectable vision models, supporting PVT v2 (ImageNet-1k) and CLIP zero-shot classification. - Detection: Bounding box object detection using RF-DETR. - Segmentation: Supports both instance and semantic segmentation. Links to the official Core AI Models and Core AI Model Zoo were also provided for further experimentation.

Related event: Developer Showcases On-Device Real-Time Vision AI on iPhone(2 posts)→

Original post →

More from Infra

Infra channel →