A YOLO dataset labeler built for capture-card and live-video workflows -- not generic photo sets. Live dual-model compare, temporal frame propagation, and tiled inference, running entirely on your own machine.
Three steps, all on your own machine.
Stream live from a capture card or import existing footage. Frames land straight in a project, deduplicated automatically.
Auto-annotate with AI assist, propagate boxes across a clip instead of frame-by-frame, and A/B two models live to see which one actually performs better on your footage.
Export a clean YOLO-format dataset and train your own model -- or run it straight in Sentinel Core Vision if that's what you're building for.
Everything a generic labeler leaves you to do by hand.
Streams straight from a capture card or live device -- built for footage, not just static photo imports.
Run two models side by side on the same live feed and see exactly which one performs better, in real time.
Label one frame, propagate boxes forward across the clip instead of re-drawing every single frame by hand.
Slice high-res frames into tiles for auto-annotation, catching small objects full-frame inference misses.
Datasets, footage, and trained models stay local. Nothing is uploaded anywhere to use the tool.
First launch downloads and configures the AI runtime automatically -- no separate Python setup required.
Cancel any time. Already-labeled datasets and exports stay yours, even if a subscription lapses.