Try our AI live in the browser: People Detection and Familiar Face Identification

World’s most accurate people detection for the edge. Try it live!

Plumerai delivers a complete software solution for people detection. Our models are tiny and fast. They even run on $1 microcontrollers. Trained with 32 million images, our AI detects people with a high accuracy in any environment.

SINGLE CORE ARM CORTEX-A

55 frames⁄s

WITH A TINY FOOTPRINT

1 MB

Functionality

  • Detects each person in view, even if partially occluded.
  • Tracks people and assigns up to 20 unique IDs.
  • Indoor, outdoor, NIR lighting.
  • Detection distance of more than 20m / 65ft.
  • Trained with 32 million labeled images.
  • Extensively validated on diverse people and settings.
  • Supports lenses up to 180 FOV.

Deploy anywhere. Run on any device, with minimal resources.

Arm Cortex-A72 @ 1.5 GHz

  • Broadcom BCM2711
  • Latency: 18.1 ms (single core)
  • Binary size: 2.4 MiB total

Arm Cortex-A53 @ 1.2 GHz

  • Texas Instruments AM62x
  • Latency: 40.3 ms (single core)
  • Binary size: 2.4 MiB total

Arm Cortex-M7 @ 280 MHz

  • ST STM32H7B3 MCU
  • Latency: 418 ms
  • Peak RAM usage: 310 KiB
  • Binary size: 1194 KiB total

Arm Cortex-M + NPU @ 500MHz

  • Realtek Ameba Pro2 SOC
  • Latency: <100 ms

Arm Cortex-M85 @ 480 MHz

  • Renesas RA8 MCU
  • Latency: 73.5 ms

Arm Ethos-U55 @ 400 MHz

  • Alif Semiconductor Ensemble
  • Latency: 12.0 ms

Processor support

  • Arm Cortex-A with/without Neon
    • Cortex-A5, A7, A9, A53, A72, etc
    • Snapdragon, NXP i.MX, TI Sitara
    • Ambarella, SigmaStar, Realtek
    • Raspberry Pi
  • Arm Cortex-M with/without Helium
    • Arm Cortex-M33, M4, M7, M55, M85
  • Arm Ethos-U55/U65
  • x86 SSE2, SSE4, AVX, AVX2, AVX512
    • Intel Celeron, i3, i5, i7, Xeon, etc
    • AMD Opteron, Ryzen, Epyc, etc
  • ESP32-S3, RISC-V, ARC EM, ARM7, ARM9
  • Various NPUs and accelerators

Operating systems

  • Linux: 32-bit and 64-bit
  • Windows: x86 and Arm
  • MacOS: x86 and Arm
  • RTOS: FreeRTOS, more
  • Bare metal

Library APIs

  • C++
  • C
  • Python

Get access to the Plumerai People Detection library.

Integration into your camera is easy with our Python, C, and C++ APIs.

Thank you for submitting the form. We’ll get in touch as soon as possible.

There was an error submitting the form. Please try again later.

Video doorbells and smart home cameras

Smart home cameras traditionally rely on motion sensors that trigger false alerts due to moving objects or lighting changes. Our accurate people detection AI removes these false alarms and enables new use cases. Receive notifications when people are on your property and not when a pedestrian simply passes by. Since all the AI runs inside the camera, there's no cloud compute cost either.

Security cameras and video analytics

Our people detection AI software runs on any standard video security camera. Since our AI is tiny, there’s no need for a GPU or AI accelerator. Our people detection enables applications such as alarms, supports people tracking for counting, loitering detection, or to steer pan-tilt-zoom cameras to auto-follow. People can even be blurred for privacy and GDPR compliance.

Video conferencing and webcams

Our people detection AI on webcams and video conferencing systems ensures that every video call participant is optimally in view. Whether you're sitting right next to the camera or at the end of a long boardroom table, you'll be visible. Automatic reports can be generated with participant statistics and conference room occupancy. Since our AI has a tiny footprint, even consumer webcams can offer professional features.

Intelligent people detection sensors

Our AI can integrate into systems that don't record or transmit video. These devices simply work better when they know where we are: lights turn on when we get home and AC's steer the cold airflow toward you. The elderly can stay independent longer with sensors that notice when they need help. Since our AI can even run on microcontrollers, these systems can be inexpensive and battery-powered.

There are many more applications where detecting people provides value. We’re also helping robots to detect humans, PIR-based motion sensors to become more accurate, detecting pedestrians to increase the efficiency of traffic lights and elevators, and many more.