Security Monitor with Image Recognition Tech

GIGABYTE, TAIPEI TAIWAN

​2018

Project Overview

This is a security monitoring software designed to deploy quickly detecting unfavoured actions in the monitored area.

Research & Interview

In the beginning, the task was to provide an alert system detecting the attempt of suicide by jumping off the rooftop, but I found we could do much more than that.During the workshop brainstorming with our algorithm engineers, I learned that the server could recognise many more specific objects and actions. Also, I see the excellent adaptation of input recourses as a great business opportunity: whoever has a security camera can apply for this product and have an alert system.

Challenge

The developers found a significant issue with image quality and speed of streaming. I unexpectedly ended up playing the critical coordinator and solved an urgent technical problem. I learned the idea of Region of Interest from Apple WWDC and proposed it to the project manager. This proposal brought the developer team extra time to fix the issue and significantly improved the user experience.

Creation process

After a few days of testing with the hardware engineer, we confirmed that enabling our user to set up the region of interest (ROI) can significantly increase image recognition efficiency.I brought the mockup created in the prototype phase to the test phase to examine our assumption. In this process, I gather insights from usability testing, revise them, and retest them.

User Interface

The service has two different levels of monitoring: camera and ROI. Users can pick up to four cameras and display live streaming on the dashboard. The product provided three alert types in this version. The number of ROI on each camera is unlimited.After we had both the structure and the wireframes confirmed, I put it all together and designed the final UI. The system comprises various card components, so it can easily and rapidly adapt to desktop and mobile devices. As a result, our security guards can receive alerts from their phones, so they will not miss any notifications while patrolling.

‍