Object Detection
EfficientDet-Lite0/2 with per-class confidence scores and bounding-box tracking.
- 80 COCO classes
- Custom
.tfliteupload - Match by photo
Benzo Vision turns your webcam into a real-time perception engine — object detection, pose estimation, hand tracking, gesture recognition and face landmarks. Save faces by name, match objects from photos, and everything stays on your device.
Each task shares the same capture pipeline, GPU delegate and overlay renderer. Save faces by name. Match objects from a photo.
EfficientDet-Lite0/2 with per-class confidence scores and bounding-box tracking.
.tflite upload33 3D landmarks with visibility scores. Compute joint angles, rep counts and form feedback.
21 landmarks per hand, up to two hands, with left/right handedness. Drives the touchless cursor.
Save faces by name. Recognition runs locally — matched people are labelled in real time.
Upload a photo of an object, name it, and every matching detection is tagged with that name.
Record the annotated canvas to MP4 (mobile) or WebM (desktop), or grab PNG snapshots.
The capture loop never blocks the UI thread. Detection runs at a fixed cadence while overlays redraw on the compositor clock.
getUserMedia / getDisplayMedia / file input feeds a single <video>.
MediaPipe Tasks runs at a fixed ~20 Hz cadence on the GPU delegate.
Signatures are compared against your saved faces and objects in under a millisecond.
Overlays are drawn on the compositor clock with cheap double-strokes, not blurs.
The composited canvas is captured at a fixed 30 Hz for smooth playback.
Models, faces and objects all live in IndexedDB. Nothing leaves your device.
No sign-up, no server, no telemetry. Open the studio and start detecting in seconds.