Project overview
Reuse the eufy multi-camera system and HomeBase edge compute already in the home, and let the first screen state a verifiable conclusion — “everything is fine at home” — while allowing “I am not sure” when confidence is low. No disease labels, no face recognition.
- Anker Hackathon · Smart security, scenario innovation
- On-device inference · judgment stays on the home network
- Four-state conclusion card · a verifiable “all is well”
- Cross-camera handoff · one target tracked across cameras
- An explicit “I am not sure” when confidence is low
Project details & context
- Event
- Anker Hackathon · Smart-security track
- Team
- Team Strawberry (3 members)
- Hardware
- eufy multi-camera + HomeBase
- My role
- Stream capture / On-device inference / Decision service / Cross-camera handoff
From a question to an experience
Users do not want “motion detected.” They want “I know everything is fine at home.” The project lets cameras already in the home deliver a verifiable conclusion instead of another alert.
A team entry in the Anker Hackathon smart-security track. It reuses the eufy multi-camera system and HomeBase edge compute already in the home, presenting a conclusion card on the first screen. I owned the hardware and inference side: stream capture, on-device inference, the decision service, and cross-camera handoff.
Keep judgment inside the home network
Capture and inference run on HomeBase and the cameras, and conclusions are generated locally. No disease labels, no face recognition — less private data leaves the home.
Give conclusions a confidence level
The conclusion card renders four states: when confident it says “everything is fine at home,” and when not it says “I am not sure,” rather than pushing a vague judgment back to the user.
Stay continuous across cameras
A single target is handed off between cameras, so a pet leaving one frame is not immediately lost and conclusions stay continuous over time.
Features · scenarios · needs · value
Core features
- On-device conclusion card: the home screen states “everything is fine at home” or “I am not sure” instead of a motion alert
- Multi-camera capture and HomeBase edge inference: judgment completes inside the home network
- Cross-camera handoff: a single target is tracked across cameras and is not lost between frames
- Confidence mechanism: an explicit “I am not sure” when the system is not sure
- Privacy by design: no disease labels, no face recognition
Application scenarios
- Pet care while at work or traveling: confirm the cat or dog is fine remotely
- Homes with several eufy cameras: living room, balcony, feeding area
- Everyday reassurance about elderly or young family members — explicitly not medical diagnosis
- Quick outings: a glance that confirms the home is all right
Needs addressed
- Users want “I know everything is fine at home,” not “motion detected”
- Alert noise and false positives train users to ignore notifications — they need conclusions
- Cloud recognition raises privacy concerns; edge inference keeps data at home
- Cameras work in isolation, with no continuous state across screens
Commercial value
- Reuses installed hardware: no new camera purchases, activating sold devices and HomeBase compute
- Shifts value from hardware margin to service: the conclusion card supports a subscription narrative
- Differentiation through scenario innovation: a verifiable conclusion versus the motion-alert red ocean
- Privacy as a selling point: on-device processing fits the sensitivity of home security data
My contribution
Team project · Edge hardware & inference
Oct 2026
- Owned eufy multi-camera stream capture and the HomeBase on-device inference pipeline, so the conclusion card is decided entirely inside the home network.
- Built the decision service and cross-camera handoff: a single target is tracked across cameras, and conclusions carry confidence — “I am not sure” when it is not sure.
Delivery & outcome
Anker Hackathon entry in the smart-security scenario-innovation track.
Continue exploring
Project archive · Original in Chinese
