Sensing layer
OmniVision and other CMOS sensor bring-up, plus TOF / optical-flow / stereo modules—so doorways and rooms stay clear and stable.
For video doorbells, peepholes, smart IPC, and home security, Shunter delivers on-device AI with a coordinated sensor–ISP–SoC stack—enabling face recognition, human detection, package recognition, and spatial sensing.
We do more than ship algorithm demos: imaging pipeline, board-level engineering, and model deployment are one trial-production-ready, regressable home-vision delivery chain.
System flow
From photons to events: the full home on-device imaging chain
A mass-production smart-home vision pipeline, split into six coordinated stages by signal and compute flow.
-
01
Sensor
OV CMOS
Optional TOF / optical flow -
02
ISP imaging
AWB / AE / HDR
Denoise and color baseline -
03
SoC / NPU
Encode, preview
On-device inference compute -
04
AI models
YOLO / MobileNet
Quantization and on-board deployment -
05
Event policy
Human / package / intrusion
Triggers and power scheduling -
06
Connect & deliver
Wi-Fi / streaming / App
Trial production and mass-production regression
Solution stack
Four layers working together
Imaging layer
ISP scene parameters (AWB / AE / HDR / denoise) for an OEM-grade home imaging quality baseline.
Intelligence layer
On-device YOLO / MobileNet models: training, quantization, NPU deployment, and power scheduling.
Product layer
Doorbell, peephole, IPC modules and boards—structure, firmware, and trial production in one delivery.
Implementation path
Six steps from requirements to mass production
Typical home-vision projects advance by stage—shortening the gap between “the algorithm runs” and “hardware ships.”
-
1
Scenario and metrics
Define detection classes, false-alarm rate, latency, power, and resolution; lock doorbell / peephole / IPC as the target form.
-
2
Sensor and platform selection
OV selection matched to SoC / NPU compute and MIPI bandwidth; add TOF / optical flow when needed.
-
3
PCBA and ISP bring-up
Schematic / layout and imaging-parameter calibration for a reproducible imaging baseline.
-
4
Algorithm training and quantization
Train detection models on home-scene data; INT8 quantize and adapt operators for the board.
-
5
Module bring-up and trial production
Firmware, event policy, wireless, and App integration; small-batch trial production for consistency.
-
6
Mass-production regression and iteration
Version control, production sampling, and ongoing model upgrades—with overseas / domestic dual delivery windows.
Case study
Featured product: Smart Video Doorbell SHT-H301
The full chain lands as a mass-production home doorbell—sensing + imaging + on-device AI in one closed loop.
Product capabilities
- Smart video doorbell solution with fingerprint / PIN and on-device face unlock
- On-device face detection / matching to reduce cloud dependency and false-unlock risk
- Pairs with peephole, Wi-Fi, and App remote management for residential entry
How this solution lands on the product
- Sensor + ISP: HDR and low-light parameters tuned for doorway backlight / night conditions
- On-device model: Lightweight face model on board; network only on event trigger
- Solution delivery: Mechanical fit, firmware integration, trial production through small batch
Customer value
- Shorten the path from demo to a mass-production solution
- Imaging and algorithms co-tuned by one team—fewer multi-vendor handoffs
- Extend to peephole, IPC, and other same-family home product lines
More scenarios
Where the same home-vision chain extends
Security IPC
Zone intrusion, human tracking, and multi-stream preview—reuse doorbell imaging and model assets.
Video peephole
Compact structure and low-power event policy, delivered from the same doorbell family.
Home security linkage
Detect–alert–record as one chain for multi-device collaboration.
Request a consultation Explore Industrial Vision Explore Robotics