Product Design
Camera Health Check
May 2021
Team: 1 full-stack engineer / 1 product manager / 1 designer
Status: Live in production (Launched in Q3 2021)
Blindsided by Broken Cameras 🕶️
Imagine this: you rely on security cameras to keep your eye on things, but a critical incident strikes. You pull up the footage, heart pounding, only to discover... a blurry mess or a view of the sky. Your camera is compromised, and you missed everything.
This scenario keeps many security professionals up at night. But what if you could detect camera issues before disaster strikes?
“Sorry, nothing is recorded because the camera was working.”
Problems
- Broken Camera: Complete camera failure renders footage unavailable.
- Tampering: Malicious actors might physically damage the camera or disrupt its functionality.
- Misalignment: Gradual or intentional shifts in the camera angle can cause it to miss crucial areas.
These issues are difficult to detect manually, especially when changes are subtle or gradual, such as a slight vibration-induced tilt in construction sites.
Factors that impacted design decisions
Alert Display: In-Platform Focus
To avoid overwhelming users, we decided to display Camera Health Check alerts within the Calipsa platform rather than through API integrations.
Unlike urgent alarms, camera health issues require less immediate attention. This approach keeps alerts organised and prevents distractions.
Technical Constraints: Motion Dependency
Ideally, Camera Health Checks would run on a regular schedule. However, due to technical limitations, we currently rely on camera motion detection to trigger checks. This dependency has impacted the initial feature design.
Setup flow
Post-launch enhancements
Day and Night Challenges
Our initial ML model struggled to differentiate between day and night images due to lighting variations. To address this, we implemented a system to automatically capture reference images for both day and night conditions. This approach improves the model's accuracy by providing specific data for comparison.
Autoplay mode
Initial user feedback revealed challenges in detecting subtle angle changes even when comparing images side-by-side. To address this, we introduced an Autoplay mode. This feature overlays the detected image and reference image, alternating every 3 seconds. The dynamic comparison aids users in identifying even the most minor discrepancies.
Health Check v.2 : Customizable Check Intervals
Two years after the initial launch, we introduced a significant upgrade: Health Check v2.
Building upon improved AI analytics, this version allowed users to define custom check intervals. Previously, the system relied on camera motion detection and performed two checks daily. This new flexibility required substantial engineering modifications to accommodate user-specified frequencies.
Global settings for Health Check
Detect Pro’s status on the camera page
Streamlined dashboard with the new ‘Timeout’ tab