The Xovis PF-Series (People Flow) is the next generation of people counting sensors — designed for large-area coverage, advanced analytics, and minimal installation effort. It is the recommended sensor for any new deployment where ceiling heights allow 2–6m mounting.
Why PF over PC Series?
Coverage | Up to 8× larger per sensor vs PC-Series. Single PF-L covers 100m² at 3.5m vs ~18m² for a PC2SE at the same height. |
Processing power | Nearly 10× faster processing. 7.9 TOPS CNN-capable AI-native NPU. 44.8 GFLOPS GPU. Quad-core 4×2.1 GHz CPU. |
Installation time | No floor mask, no illumination mask, no taboo mask needed. Auto-height detection. Saves ~10 minutes per sensor on large deployments. |
Multisensor scale | Up to 15 PF sensors in a multisensor without a SPIDER. Can combine PF and PC sensors in the same multisensor. |
Attention Metrics | On-sensor visitor engagement measurement — a PF-exclusive feature. Built on View Direction AI, measures how long visitors look at displays, products, or signage. |
Firmware | Runs Firmware 5.9+. Same API, same HUB interface, same REST endpoints as PC Series — no additional integration effort. |
Note: Staff Exclusion is currently NOT available on PF-Series. New approaches are in development. If staff exclusion is a hard requirement for a project, use PC-Series sensors for that entrance.
Technical highlights
Chip | Synaptics next-generation SoC — AI-native NPU architecture |
NPU performance | Up to 7.9 TOPS CNN — 22.5× faster than PC2SE, 3× faster than NXP i.MX8 |
Mounting height | 2.0 – 6.0 m (PF-L model) |
Coverage at 2.5m | 5.60 × 5.60 m = 31.36 m² |
Coverage at 3.5m | 10.0 × 10.0 m = 100 m² |
Max multisensor (no SPIDER) | Up to 15 sensors |
Max multisensor (with SPIDER) | Up to 350 sensors (SPIDER) |
Power | PoE |
Privacy | GDPR-compliant, fully compatible with evolving privacy regulations |
Firmware | FW 5.9+ (FW 5.10 coming with staff analytics and enhanced Attention Metrics) |
AI extensions available on PF-Series
Attention Metrics | PF-exclusive. Measures visitor gaze/engagement at displays. Powered by View Direction AI. Ideal for retail marketing ROI measurement. |
Gender statistics | Available. Improved AI model for gender estimation in FW 5.9.2. |
Apparent age estimation | Available. Configurable age bins (e.g. 18–28, 29–39). Age histogram logic added in FW 5.9.2. |
View direction | Available |
Object detection | Wheelchair, pram/stroller, bicycle, shopping cart — all available |
Staff exclusion | NOT yet available on PF-Series. Coming in future firmware. |
What's new in recent firmware
FW 5.9 (Feb 2026) | Baseline PF-Series release. New stitching algorithm for multisensors. Visualization maps in live view. Recording Viewer enhancements. |
FW 5.9.1 (Mar 2026) | Stability improvements and performance fixes for PF, PC, and SPIDER. |
FW 5.9.2 (May 2026) | Custom logic with age histograms. Improved AI model for gender statistics and apparent age on PF-L. Stability improvements. |
FW 5.10 (coming soon) | Staff analytics (behaviour-based staff exclusion via defined zones/lines). New Attention Metrics enhancements. |
Connecting to Vemcount
PF-Series sensors connect to Vemcount exactly the same way as PC-Series — via PoE data push to Vemcount's ingestion endpoint. The sensor MAC address is registered in Settings → Devices. All counting lines, zones, and extension data appear automatically in Vemcount once the sensor is online and configured in Xovis HUB Device Control.
Recommended use cases: Large retail stores, shopping malls, airports, transit concourses, open-plan offices. Any location where fewer sensors per floor area is a cost driver. PF-L at 3.5m ceiling covers 100m² with a single sensor — a typical 300m² store entrance needs just 3 sensors vs 15+ PC2SE sensors.