Roadmap

What's shipped, what we're building, and what's planned. We only list what we're confident about.

Shipped · 1
Scene Text Recognition
Perception — single-frame win
In progress · 9
80 everyday object types, out of the box (COCO-80)
Perception — single-frame win
Detection
Perception — single-frame win
Segmentation
Perception — single-frame win
Depth
Perception — single-frame win
Basic tracking
Perception — single-frame win
One clean box per detection without tuning — Batch 3ms (NMS-free)
Batch parity, breadth, text
Oriented boxes (OBB)
Batch parity, breadth, text
Heading / rotation detection
Batch parity, breadth, text
Zones and lines — in-zone count + line-cross count
Density + the spatial foundation
Planned · 19
Pose
Batch parity, breadth, text
Keypoints on every class
Batch parity, breadth, text
Runs on Intel chips, not just NVIDIA (OpenVINO)
Batch parity, breadth, text
Behavior
Video understanding + identity
Motion (optical flow)
Video understanding + identity
Respiration
Video understanding + identity
Heart rate
Video understanding + identity
Recognition across cameras (ReID)
Video understanding + identity
Counting (density estimation)
Density + the spatial foundation
Proximity alerts
The event core (“alert me when”)
Zone enter/exit
The event core (“alert me when”)
Count-in-region (incl. crowd-density-in-zone)
The event core (“alert me when”)
Fires once when it matters (debounce, hysteresis)
The event core (“alert me when”)
Notifies your code directly (event queue + callbacks)
The event core (“alert me when”)
Dwell, appear/disappear
Relational events + verticals
Threshold alerts (speed over X, count over N)
Relational events + verticals
Possession transfer — who is holding what, and when it changes hands
Relational events + verticals
Ag expansion pack
Relational events + verticals
DeepStream integration — one-line add
Deployable
Want something that isn't here? Request a feature →