Features

Speak or type a command and it is planned into a chain of system actions and executed.

On-device speech-to-text (Turkish + English) and spoken replies — no cloud.

The camera narrates itself aloud, as a subtitle — watch it or just listen.

Lock a subject and the digital PTZ keeps them centred as they move.

See where a tracked subject is even while they are behind cover.

See each person's attention direction and who is interacting with whom.

Draw a box on any region and get a crisp, photographic close-up.

People, vehicles, animals and hazards detected on every analysed frame.

Per-pixel foreground / object masks for cleaner geometry and analytics.

Dense monocular depth from a single ordinary camera, no stereo rig.

Depth lifted toward real-world units for distance and size readouts.

Vehicle body type, brand, colour; person clothing colour, stature and behaviour.

Stable identities across frames, including brief occlusions.

Dozens of simultaneous tracklets with per-track motion state.

Recovers camera motion so speeds are measured against the ground.

Optional intrinsics + ground-plane calibration for metric accuracy.

A structured read of the scene: entities, ground, background, hazards.

Reason about position, distance, convergence and near-future motion.

Turn a frame into a navigable 3D point cloud you can fly through.

Capture a few seconds of the scene in 3D and replay it as a video you can fly through.

Coloured point clouds lifted from depth + RGB, per camera.

Triangulated surfaces from the depth field for solid scene geometry.

Fills and textures the scene hidden behind people and vehicles.

Real-time counts, density, motion, speed and per-camera reputation.

Zones, rules and correlated threat scoring raise the right alerts.

Cross-camera re-identification, dossiers and relationship graphs.

Analyse a single still: detect, segment, depth, lift to 3D.

Batch-analyse recorded video into events, tracks and 3D scenes.

Runs on a single local workstation, no cloud round-trips.

CUDA + FP16 inference across detection, depth and super-resolution.

Fully air-gapped: no external calls, all data stays on the host.