CoreAIKit

ActionCamera

Live action recognition from the camera, fully on-device: CameraFeed streams frames, a rolling 16-frame clip goes through ActionRecognizer (V-JEPA 2 ViT-L, SSv2 head — Meta’s video world model) and the top actions come back with confidences.

The whole ML surface is two calls:

let recognizer = try await ActionRecognizer(catalog: "vjepa2-vitl-ssv2")
let actions = try await recognizer.classify(videoAt: clipURL)   // or classify(frames:)

Run

xcodegen generate
open ActionCamera.xcodeproj

Run on an iPhone (camera required). First launch downloads the model from the Hugging Face Hub; later launches load from cache.

CLI (macOS, no Xcode)

swift run action-cli --video sample.mp4

sample.mp4 is a synthetic clip (a hand pushing a block from left to right) so the command works out of the box; point --video at a real clip for real footage.

This site is open source. Improve this page.