BAV ONE Computer Vision Operating System · Turn existing cameras into programmable AI sensors
Explore the platformPerson detection provides a person class and bounding box for configured video scenes; it is not facial recognition and does not establish identity.
How it works
Each stage remains observable and can be validated independently.
An approved model evaluates frames and returns a person class, confidence and location. Thresholds are selected against representative scenes rather than assumed globally.
Angle, scale, occlusion, lighting, motion and image quality affect detection. Within-camera tracking supports movement, direction and dwell without claiming identity.
Combine a person track with configured polygons, boundaries, schedules and duration before creating an event.
Use assessed person detection for campus, factory, warehouse and perimeter workflows with appropriate privacy and human review.
Capability, compatibility and performance depend on the camera scene, approved models, infrastructure and acceptance criteria. Human review remains appropriate for critical decisions.
Related guidance
Frequently asked questions
No. Person detection locates the person class in a frame. It does not identify who the person is.
Yes. BAV ONE supports assessed edge, on-premises, cloud and hybrid architectures. The selected design depends on camera access, compute, bandwidth, data-control and operational requirements.
No. Compatibility depends on stream access, protocol, resolution, frame rate, camera angle, lighting, network conditions and the selected use case. Bharat AI Vision validates these factors before rollout.
Begin with a camera and workflow assessment, then validate a focused one-to-four-camera pilot against agreed operating conditions and acceptance criteria.
Validate before rollout
Map the operating requirement, assess compatible streams and validate a focused workflow before expanding.
